Deliberative and non-deliberative effects of descriptive and injunctive norms on cancer screening behaviors among African Americans

Abstract

Objective:

Two longitudinal studies examined whether effects of subjective norms on secondary cancer prevention behaviors were stronger and more likely to non-deliberative (i.e., partially independent of behavioral intentions) for African Americans (AAs) compared to European Americans (EAs), and whether the effects were moderated by racial identity.

Design:

Study 1 examined between-race differences in predictors of physician communication following receipt of notifications about breast density. Study 2 examined predictors of prostate cancer screening among AA men who had not been previously screened.

Main Outcome Measures:

Participants’ injunctive and descriptive normative perceptions; racial identity (Study 2); self-reported physician communication (Study 1) and PSA testing (Study 2) behaviors at follow up.

Results:

In Study 1, subjective norms were significantly associated with behaviors for AAs, but not for EAs. Moreover, there were significant non-deliberative effects of norms for AAs. In Study 2, there was further evidence of non-deliberative effects of subjective norms for AAs. Non-deliberative effects of descriptive norms were stronger for AAs who more strongly identified with their racial group.

Conclusion:

Subjective norms, effects of which are non-deliberative and heightened by racial identity, may be a uniquely robust predictor of secondary cancer prevention behaviors for AAs. Implications for targeted screening interventions are discussed.

In the United States, African Americans (AAs) generally have the highest cancer incidence and mortality rates (DeSantis et al., 2016; O’Keefe, Meltzer, & Bethea, 2015; Siegel, Miller, & Jemal, 2017). The multifaceted correlates of these disparities - e.g., disadvantaged socioeconomic factors, worse access to health care, lower satisfaction with care (DeNavas-Walt & Proctor, 2015; Lurie, Zhan, Sangl, Bierman, & Sekscenski, 2003; Tarraf, Jensen, & González, 2017) – are also associated with differences in cancer screening rates which ultimately contribute to disparities in incidence and mortality across a range of cancer sites (Ademuyiwa et al., 2011; The Center to Reduce Cancer Health Disparities, 2008). Compelling evidence shows that cancer screening deficits lead to worse incidence and mortality outcomes for AAs for breast (Curtis, Quale, Haggstrom, & Smith-Bindman, 2008; Smith-Bindman et al., 2006; van Ravesteyn et al., 2011), prostate (Carpenter et al., 2010; Jones et al., 2008), and colorectal cancers (Lansdorp-Vogelaar et al., 2012). Hence, facilitating cancer screening among AAs is a critical means to address cancer health disparities. A greater understanding of psychological determinants of cancer screening behaviors among AAs can illuminate mechanisms that may be employed to facilitate cancer screening uptake. This current research examined how AAs’ perceptions of behavioral norms influenced cancer screening decision-making, and demonstrated that the effects of norms may be partially independent of AAs’ behavioral intentions (i.e., non-deliberative), and that racial identity moderated the effects norms on AAs’ cancer screening behaviors.

Subjective Norms and Behavior

The disheartening data about deficits in cancer screening rates among AAs, and related media narratives, likely influence subjective perceptions of behavioral norms (i.e., subjective norms) for African Americans. Extant literature supports the influence of subjective norms on subsequent behaviors (e.g., Ajzen, 1991; Cialdini, 2003; DeJong, 2010; Lapinski & Rimal, 2005), and recent research shows how subjective norms are related to preventive health behaviors such as condom use and sun protective behaviors (Lewis, Litt, Cronce, Blayney, & Gilmore, 2014; Reid & Aiken, 2013). For AAs’, compared to European Americans (EAs), subjective norms may matter more for health behavior decision-making. For example, subjective norms were the strongest predictor of breast feeding intentions for AA mothers (Bai, Wunderlich, & Fly, 2011). Compared to EAs, subjective norms were stronger predictors of teenagers’ smoking intentions (Hanson, 1997) and fruit and vegetable intake (Blanchard et al., 2009) for AAs. These findings suggest that subjective norms may play a similarly amplified role among AAs in the context of cancer screening behavioral decision-making.

Psychological Mechanism Relating Norms to Behavior

The role of subjective norms in behavioral decision-making is instrumental in contemporary behavioral prediction models. One such model, the Theory of Planned Behavior (TPB: Ajzen, 1991, 2011), provides the theoretical framework we used for this research. Briefly, TPB posits that attitudes, subjective norms and perceived behavioral control influence behavioral intentions, which in turn influences actual behavior; hence, subjective norms are indirectly related to behaviors via behavioral intentions. Consistent with conceptualization of subjective norm in most decision-making theories (e.g., Cialdini, 2003; Lapinski & Rimal, 2005), TPB distinguishes two types of subjective norms: descriptive norms are individuals’ perceptions of what relevant others do, and injunctive norms are individuals’ perceptions of what relevant others want you to do (Fishbein & Ajzen, 2010; Manning, 2009, 2011a, 2011b; McMillan & Conner, 2003; Sheeran & Orbell, 1999). The TPB has been used successfully to predict cancer screening behaviors (Griva, Anagnostopoulos, & Madoglou, 2009; Jennings-Dozier, 1999; Sieverding, Matterne, & Ciccarello, 2010; Tolma, Reininger, Evans, & Ureda, 2006), and is appropriate for examining the unique roles of injunctive and descriptive norms (collectively, subjective norms) in cancer screening behavioral decision-making among AAs.

Stronger Effects of Subjective Norms For African Americans

Our expectations of an amplified role of subjective norms for AAs are informed by social identity theory (SIT: Tajfel & Turner, 2004), the effects of racial identity salience, and the effects of greater collective self-concept. SIT posits that one’s group membership (i.e., social identity) influences her/his cognitions and behaviors. SIT-derived hypotheses, that people prefer in-group members whose behaviors are consistent with group norms (Hogg & Abrams, 1988; Hogg & Turner, 1987), have been empirically supported (e.g., Bettencourt et al., 2016; Marques, Abrams, Paez, & Martinez-Taboada, 1998; Marques & Yzerbyt, 1988; Marques, Yzerbyt, & Leyens, 1988). Relatedly, evidence shows that norms have stronger effects on behavior when a relevant identity is made salient (Byungjoo, Seungjun, & SangHyun, 2017; Reicher, 1984; Wellen, Hogg, & Terry, 1998); and, experiences of discrimination (Sanders Thompson, 1999), as well as contextual factors (Forehand, Deshpandé, & Reed Ii, 2002; Kenny & Briner, 2013; Shelton & Sellers, 2000; Steck, Heckert, & Heckert, 2003), can make racial identity more salient. Experiences of racial discrimination and race-based medical mistrust in contexts of medical interactions generally (Penner, Albrecht, Coleman, & Norton, 2007; Penner et al., 2009), and in cancer-related interactions specifically (Penner et al., 2012; Thompson, Valdimarsdottir, Jandorf, & Redd, 2003; Thompson, Valdimarsdottir, Winkel, Jandorf, & Redd, 2004), can heighten the salience of racial identity when AAs make decisions related to cancer screening. Finally, when individuals have more collectivist concept of the self (i.e., relatively greater weight of group membership in defining one’s self concept), subjective norms more will more strongly influence behavioral decisions (Trafimow & Finlay, 1996; Ybarra & Trafimow, 1998). Studies have shown that AAs are more collective in their self-concepts and afford greater centrality to ethnic identity compared to EAs (Avery, Tonidandel, Thomas, Johnson, & Mack, 2007; Gaines Jr., Larbie, Patel, Pereira, & Sereke-Melake, 2005; Gaines Jr., Marelich, Bledsoe, & Steers, 1997; Kern & Grandey, 2009; Marshall & Naumann, 2018). Altogether, these conditions are apt for subjective norms to have stronger effects on AAs’, compared to EAs’, screening decision-making.

The distinction between group norms as conceptualized by SIT (i.e., perceptions of normative behaviors for a group represented in one’s identity schema) and subjective norms as conceptualized by TPB (injunctive and descriptive behavioral perceptions of relevant others) warrants clarification. Typically, assessment of subjective norms prescribed by the TPB hold that salient referents influence normative perceptions. Depending on the behavior, salient identity-relevant groups can inform subjective norms and subsequent behaviors. For example, in a study of alcohol consumption behaviors among university military and veteran students, normative perceptions related to “typical students” influenced drinking behaviors whereas normative perceptions related to students in the military did not (Miller et al., 2016), suggesting that relevant others may be conceptualized as broader in-group.

Non-Deliberative Effect of Subjective Norms

TPB posits an indirect effect of injunctive and descriptive norms; however, some research provides evidence for a non-deliberative effect of norms on behavior – that is, a residual direct effect that is unmediated by deliberative behavioral intentions (Manning, 2009, 2011a, 2011b). This non-deliberative effect represents a more impulsive/automatic response to normative information, in contrast to a more reflective/conscious response (see Strack & Deutsch, 2004). We distinguish this conceptualization from prior research in that we focus on behaviors that are determined by deliberative intent, whereas most others have examined automaticity in the relation between normative perceptions and behaviors that were not determined by deliberative choice (Aarts & Dijksterhuis, 2003; Bargh & Chartrand, 1999; Chartrand & Bargh, 1999; Dijksterhuis & Bargh, 2001) – see also Manning (2011a - footnote 1). A recent meta-analysis supported the non-deliberative effect of descriptive norms on health behaviors (McEachan et al., 2016). In that effects of subjective norms are often under detected (Nolan, Schultz, Cialdini, Goldstein, & Griskevicius, 2008), and thus arguably occur at least partially independent of conscious intent, we expected that heightened effects of subjective norms for AAs would manifest more via the non-deliberative route. Psychological (e.g., social motivation, self-control, effortful control, cognitive capacity) and contextual (e.g., more social behaviors) constructs have been shown to moderate the extent to which deliberative vs non-deliberative processes influenced behavior in relevant dual-process models (Grenard, Ames, & Stacy, 2013; Honkanen, Olsen, Verplanken, & Tuu, 2012; Manning, 2011a, 2011b; Pieters, Burk, Van der Vorst, Engels, & Wiers, 2014). We will examine whether racial group membership also moderates the non-deliberative effect of subjective norms.

Current Research

We examined support for our hypotheses with secondary analyses of two separate sets of data from our prior research. With the first study, we examined physician communication behaviors among women who were notified about a breast cancer risk factor. We expected that (H1) subjective norms would more strongly affect behaviors of AA women compared to EA women, and that (H2) the effect of subjective norms would be non-deliberative among AA women. With the second study, we further examined support for a non-deliberative effect of subjective norms on prostate cancer screening decision-making among AA men, and examined (H3) racial identity as a moderator of the effects of subjective norms.

Study 1: Normative Influences on Physician Communication

Data for Study 1 were from an examination of AA and EA Michigan women’s physician communication behaviors following notifications about breast density and breast cancer risk (Manning, Albrecht, O’Neill, & Purrington, 2018; Manning et al., 2017). Briefly, women with dense breasts are at increased breast cancer risk (Barlow et al., 2006; Boyd, Martin, Yaffe, & Minkin, 2011; McCormack & dos Santos Silva, 2006), and a recently adopted law in Michigan mandate that such women are notified of this on their mammogram report. These data were collected to examine between-race differences, and determinants of differences, in processes and outcomes related to subsequent physician communication following notification.

Method

Participants and procedures.

Participants were women with dense breasts, no prior breast cancer diagnoses, who presented for routine screening mammograms and whose screening results were negative (N = 557). Within two weeks of their mammograms, women participated in an initial survey investigating “opinions and perspective on some information that was included in your mammogram report”; three month later they were invited to participate in a follow-up survey assessing, among other things, whether they communicated with their physicians about the notifications. Participants completed the surveys online (Qualtrics, 2015), and received a $25 gift certificate for participation in each survey. The study was approved by Wayne State University’s Institutional Review Board.

Two hundred and seventy one women responded to the follow-up survey (49% response rate), of which there were 91 AAs (43%) and 121 EAs (57%) who comprised the sample for analysis. EA women were more likely to follow up than AA women (57% vs. 37%; χ²(1) = 19.42, p < .01); consequently, responders were higher income (d = 0.35), more educated (χ²(2) = 9.78, p < .01), and more likely to be married (χ²(2) = 8.83, p < .05) compared to non-responders. Of the analysis sample, most women (96%) women reported having health insurance. AA women were less likely married or partnered χ²(2) = 27.00, p < .01, less educated (χ²(5) = 14.29, p < .05) and had lower income ($40K to $49K vs. $90K to $99K, p < .01) than EA women.

Measures.

Relevant TPB predictors (i.e., attitudes, injunctive and descriptive norms, PBC and intentions) and sociodemographic descriptors were assessed with the initial survey, and behavior was assessed at follow-up.

TPB variables.

The behavioral target was “Talking to my doctor (i.e., primary care, ob/gyn, etc.) about the notification regarding the density of my breasts within the next three months…” Item responses were on a 7-point scale from “strongly agree” to “strongly disagree” (except for semantic differential attitude items), with some items reverse scored to address biased responding due to responder acquiescence. Two items assessed behavioral intentions (“I intend to talk…”, “I have decided that I will talk…”; r = .53, p < .001). Five semantic differential items (e.g., “Pleasant – unpleasant”, “Extremely good – extremely bad”; α = .83) assessed behavioral attitudes. Perceived behavioral control (PBC) was assessed with the item “Whether or not I talk… Injunctive norms were assessed with two items (e.g., “Most people I care about would expect me to talk…”, “Most people who are important to me would want me to talk…”; r = .58, p < .001). Descriptive norms were assessed with two items (e.g., “Most women who I care about would talk…”, “I could see most of the women who are important to me talking…”; r = .45, p < .001).” Behavior was assessed with Yes-No responses to the item “In the past three months, have you talked to your doctor (your primary care doctor, obstetrician/gynecologist, or whichever doctor you discuss women’s health with in particular) about the breast density notification you received with your last mammogram report.”

Prior BD Awareness.

We controlled for women’s prior awareness of their BD, which was assessed with Yes-No responses to the item “Prior to receiving your current mammogram report, did you know how dense your own breasts are?”

Planned analysis.

We fit a multigroup path model where racial group membership defined groups. Behavior was regressed onto prior BD awareness, intentions, PBC, and injunctive and descriptive norms (coefficients for norms represent non-deliberative effects). Intentions in turn were regressed onto attitude, PBC, injunctive and descriptive norms and prior BD awareness. We used lavaan (Rosseel, 2012) with R (R Core Team, 2016) to fit the path models. Given our relatively small sample size (Long, 1997; Moshagen & Musch, 2014), we used a linear probability model (Hellevik, 2009) fit via maximum likelihood estimation. We used chi-square difference tests to establish between-race differences in associations with behavior by comparing models in which the coefficients for the predictors of behavior were constrained to be equal (invariant model) vs freely estimated (variant model) between groups.

Results

Descriptive statistics are presented in Table 1. AA women had more favorable attitudes, injunctive and descriptive normative perceptions and behavioral intentions; EA women were more likely to report being aware of their breast density prior to receiving notification.

Table 1:

Study 1: Descriptive statistics for AA and EA Women

European Americans					Correlations							African Americans
		M	SD	N	(1)	(2)	(3)	(4)	(5)	(6)	(7)		M	SD	N
Prior BD Awareness**b	(1)	0.60c	--	121	--	−.03	−.13	−.09	.20	.07	.18	(1)	0.33c	--	90
Attitude**a	(2)	5.45	1.20	121	−.21*	--	.48**	.36**	−.05	.48**	.03	(2)	5.93	1.07	87
Injunctive Norm*a	(3)	4.79	1.47	121	−.16	.58**	--	.62**	−.07	.52**	.05	(3)	5.32	1.66	91
Descriptive Norm*a	(4)	4.90	1.22	121	−.08	.49**	.50**	--	.14	.49**	.21*	(4)	5.29	1.40	91
PBC	(5)	6.03	1.33	120	.10	.06	.01	.06	--	.27**	.00	(5)	5.90	1.37	91
Behavioral Intentions**a	(6)	5.00	1.57	121	−.10	.70**	.64**	.54**	.07	--	.05	(6)	5.63	1.37	91
Behavior	(7)	0.27c	--	121	.00	.23*	.18*	.23*	.10	.31**	--	(7)	0.34c	--	91

Notes: BD = breast density, PBC = perceived behavioral control. Correlations above diagonal are for African Americans, and below the diagonal are European Americans. Lettered superscripts a and b indicate significant mean differences indicated by t-test and chi-square respectively; superscript c = proportion.

^*= p < .05

^**= p < .01

Model comparisons indicated that the coefficients predicting behavior were significantly different between EA and AA women (Δχ²(8) = 16.97, p < .05), and fit indices indicated good overall model fit for the variant model, χ²(2) = 0.07, ns; RMSEA = 0.00, 90% CI = 0.00 – 0.00; CFI = 1.00; TLI = 1.00; SRMR < .01. As seen in Table 2, attitudes, and both injunctive and descriptive norms were significant predictors of intention for all women, and PBC was a significant predictor of intentions only for AA women. Intentions were a significant predictor of behavior for EAs, whereas it was not for AAs. In contrast, descriptive norms were a significant predictor of behavior for AA women, demonstrating a significant non-deliberative effect of descriptive norms in support of H2. The total effect of descriptive norms on behavior was significant for AA women (estimate = 0.10, p < .01), whereas neither injunctive nor descriptive norms had significant total effects on behavior for EA women, demonstrating stronger effects of descriptive norms for AA women in partial support of H1.

Table 2:

Study 1: TPB path analysis results

	EA		AA
	Coefficients	SE	Coefficients	SE
Behavior
Intention	0.08*	0.03	−0.01	0.05
PBC	0.02	0.03	−0.04	0.04
INJN	−0.01	0.04	−0.02	0.04
DN	0.03	0.04	0.10*	0.05
Prior BD Awareness	0.03	0.08	0.21†	0.11
Intention
Attitude	0.58**	0.10	0.39**	0.12
PBC	0.03	0.07	0.26**	0.08
INJN	0.33**	0.08	0.20*	0.09
DN	0.22*	0.09	0.18†	0.10
Prior BD Awareness	0.19	0.19	0.29	0.24

Notes: EA = European American, AA = African American. PBC = perceived behavioral control, INJN = injunctive norms, DN = descriptive norms, BD = breast density.

^†= p < .10

^*= p < .05

^**= p < .01.

Study 2: Racial Identity as a Moderator of Normative Influence

Prior research indicates that stronger identification with one’s in-group (i.e., group identity) is associated with greater in-group favoritism (e.g., Crisp & Beck, 2005; see Hewstone, Rubin, & Willis, 2002 for review) and stronger preferences for norm consistent behaviors among in-group members (Marques et al., 1988). Other research indicates that ethnic identity is more salient and central to AAs compared to EAs and other minority groups (Avery et al., 2007; Brown et al., 2014; Kern & Grandey, 2009; Marshall & Naumann, 2018; Utsey, Chae, Brown, & Kelly, 2002; Yoon, 2011). This heightened salience of racial identity for AAs supports our theoretical rationale for the stronger role of subjective norms on AAs’ behaviors. It also suggests that when we examine normative influence among AAs, both deliberative and non-deliberative effects of subjective norms may be stronger for AAs who more strongly identify with their racial group (H3). We examined this hypothesis in Study 2 with data from the control arm of a project examining the effects of a culturally targeted intervention on AA men’s prostate cancer screening intentions and behaviors.

Method

Participants and Procedure.

Eligible participants were English-speaking AA men between 40 and 75 years of age, with no history of prostate cancer, who lived or worked within the New York City (NYC) metropolitan area and reported no prostate-specific antigen (PSA) testing in the prior six months. Participants (N = 197 after excluding 15 due to problems with consent forms, literacy, and missing data) completed self-administered questionnaires in groups (average size = 5) at local community sites (e.g., public libraries) from February 2006 to July 2007. At baseline, consented participants completed the baseline questionnaire and were then randomly assigned to receive either a culturally targeted brochure (N = 100) or a generic brochure published by the American Urological Association titled, Prostate Cancer Awareness for Men (N = 97). Since we had no hypotheses about the effects of targeted messages, only participants who received the generic brochure were included in these analyses. After reading the brochure, participants immediately completed a post-intervention questionnaire. Typically, these sessions lasted about 90 minutes. Participants were mailed a follow-up questionnaire six months later. Sixty one participants (63% of analysis sample) completed the follow-up, of whom 56 provided behavioral data described below. Follow-up responders did not differ from non-responders on any of the measures included in analyses. Participants were compensated $50 for their participation in the entire study.

Measures.

All TPB variables (PSA screening attitudes, injunctive and descriptive norms, perceived control, and intentions) were assessed at baseline and, with the exception of injunctive and descriptive norms, were repeated post-intervention. Subjective norms were excluded from the post-intervention survey as the targets of the intervention were attitudes and perceptions of control. We used the baseline assessments of subjective norms and post-intervention assessments of other TPB variables for hypothesis testing. Racial identity was assessed at baseline, and behaviors were assessed at follow-up.

TPB Variables.

The behavioral target was “[having] a PSA in the next 6–7 months.” Responses were assessed on a scale from 1 (strongly disagree) to 5 (strongly agree). We assessed both instrumental and affective components of attitudes. The mean of participants’ responses indicating whether they thought that getting a PSA would be worthwhile, reassuring, wise, healthy and important was used to assess instrumental attitudes (α =.95). The mean of participants’ responses indicating whether they thought getting a PSA would be worrying, embarrassing, and unpleasant (all reverse scored) assessed affective attitudes (α = .82). Injunctive norms were assessed with the mean of the items, “Most people who are important to you think you should be screened for prostate cancer” and “People who are important to you have encouraged you to be screened for prostate cancer,” (r = .65, p < .01). Descriptive norms were assessed with the mean of the items “I have talked to or heard from men who have been screened for prostate cancer” and “I have talked to or heard from men who benefit from regular prostate cancer screening,” (r = .83, p < .01). Whereas these items do not directly assess perceptions of what other men do with regards to screening behavior, they may be considered reasonable proxies in that responses to the items likely reflect perceptions of whether other men get screened. Perceived behavioral control was assessed with the item, “How much control do you have over getting a PSA?” Participants responded on a scale from 1 (Complete control) to 5 (No control) – responses were reverse coded so that higher numbers represented more control. Behavioral intentions were assessed with responses to the item “I intend to have a PSA in the next 6–7 months.” At the six-month follow-up, men who indicated that they had gotten, or were scheduled for, a PSA test since baseline were coded as engaging in PSA screening behaviors.

Racial Identity.

The 8-item centrality subscale of the Multidimensional Inventory of Black Identity (MIBI: Sellers, Rowley, Chavous, Shelton, & Smith, 1997) was used to assess the extent to which race is a core component of self-concept (e.g., “In general, being Black is an important part of my self image.”). Participants responded on a scale from 1 to 4 (α = .66).

Planned analysis.

We took a similar linear probability approach to fit a path analysis with lavaan (Rosseel, 2012). We regressed intentions onto instrumental and affective attitudes; PBC; descriptive and injunctive norms; racial ID; and two-way interactions between each subjective norm and racial ID. We simultaneously regressed behaviors onto intentions, PBC, descriptive and injunctive norms, racial ID, and the two way interactions. To address potential multicollinearity and to facilitate interpretation of interactions, we used standardized values for the subjective norms and racial identity for model main effects and product interaction terms. We probed significant interactions with methods recommended by Preacher et al. (2006).

Results

Means, SDs and correlations for study variables are presented in Table 3. Despite the small sample size, the final model exhibited decent fit (χ²(2) = 3.04, ns; RMSEA = 0.10, 90% CI = 0.00,0.30; SRMR = .02; CFI = 0.94). Path coefficients are presented in Table 4. Intentions were unrelated to subsequent 6-month behavior; hence, there were no significant indirect of subjective norms, despite a significant association between injunctive norms and intentions. A significant main effect of injunctive norms on behavior once again supported the non-deliberative effect of normative information for AAs. There was also a significant descriptive norms by racial ID interaction on behavior (illustrated in Figure 1). In support of H3, probing indicated a significant positive direct effect of descriptive norms on behavior when racial ID was high (≥ 2.4 SD above mean); additionally, there was a negative direct effect when racial ID was low (≤ 2.5 SD below mean). With deference to sample size, we tentatively probed the negative direct effect of descriptive norms for evidence of a suppressor effect by regressing behavior unto descriptive norms followed by injunctive norms in a stepwise fashion among AA men who were below median racial ID (N = 23). Whereas the descriptive norm coefficient was non-significant without injunctive norms in the model (b = −0.13, p > .25), it was marginally significant when injunctive norms was introduced (b_descriptive = −0.20, p = .07; b_injunctive = 0.34, p = .02).

Table 3:

Study 2: Means, SD and correlations for study variables

		M	SD	r
				1	2	3	4	5	6	7
1	Instrumental Attitude	4.58	0.47
2	Affective Attitude	3.81	0.95	0.13
3	Descriptive Norms	3.08	1.31	0.16	0.17
4	Injunctive Norms	3.71	0.96	0.21	−0.02	0.16
5	PBC	4.43	0.99	−0.01	.45**	0.25†	−0.02
6	Intentions	3.66	1.08	.29*	0.12	0.23†	.39**	0.07
7	Racial ID	2.92	0.44	0.10	−0.16	−0.04	−0.25†	−0.16	0.01
8	PSA Behavior	0.50	0.50	0.11	0.16	0.08	.42**	−0.07	0.25†	−0.08

Notes: PBC = perceived behavioral control.

^†= p < .10

^*= p < .05

^**= p < .01.

a = proportion

Table 4:

Study 2: Path analyses results

	Coefficients	SE
Behavior
Intentions	0.04	0.06
PBC	−0.05	0.06
Descriptive Normsz	0.01	0.06
Injunctive Normz	0.23**	0.07
Racial IDz	0.02	0.06
Descriptive Norms by Racial ID	0.14*	0.06
Injunctive Norm by Racial ID	−0.04	0.06
Intentions
Instrumental Attitudes	0.37	0.29
Affective Attitudes	0.10	0.15
PBC	0.03	0.15
Descriptive Normsz	0.12	0.14
Injunctive Normz	0.44**	0.15
Racial IDz	0.14	0.14
Descriptive Norms by Racial ID	0.07	0.14
Injunctive Norm by Racial ID	−0.08	0.14

Notes: PBC = perceived behavioral control.

^**= p < .01

^*= p < .05

^†= p < .10. Subscript z = standardized.

Figure 1:

Study 2: Influence of racial identity on direct effect of descriptive norms on PSA screening behavior. Low Racial ID = −3 SD, High racial ID = 3 SD. ω₁ = simple slope.* = p < .05.

Discussion

Our data provide preliminary support for the amplified role of subjective norms in AAs behavioral decision making related to secondary cancer prevention behaviors, and is the first to show that racial group membership moderates non-deliberative effects of subjective norms. Results from Study 1 partially supported our hypothesis that subjective norms have stronger effects on behaviors for AAs compared to EAs; specifically, there was a significant total effect of descriptive norms for AA, but not for EA women. Results from both studies supported our hypothesis regarding a significant non-deliberative effect of subjective norms on behaviors for AAs. Specifically, descriptive norms directly predicted physician communication about breast density notifications for AA women, and both injunctive and descriptive norms directly predicted PSA testing behaviors for AA men. In partial support of our hypothesis regarding the effects of racial identity however, the non-deliberative effect of descriptive norms on PSA testing behaviors was only present when men more strongly identified as AAs. These findings suggest a gender difference in the non-deliberative effect of injunctive norms since there was no significant direct effect of injunctive norms on physician communication for AA women; however, future research should examine these potential gender differences among secondary prevention behaviors applicable to both men and women (e.g., colorectal cancer screening).

Racial identity did not influence the non-deliberative effects of injunctive norms; however, our hypothesis was partially supported for the non-deliberative effects of descriptive norms. An inverse effect of descriptive norms on PSA testing behaviors was noted among AA men for whom racial identity was low. Probing this effect among AA men with weaker racial ID indicated evidence of a positive suppressor effect (i.e., where the introduction of a subsequent predictor suppresses unexplained variance in the association between a prior predictor and the outcome, thus strengthening the coefficient for the prior predictor). Suppressor effects, and moderation of suppressor effects, have been documented in other research using TPB to examine the joint effects of descriptive and injunctive norms (Manning, 2009, 2011b; Manning, Wojda, et al., 2016) – we demonstrated similar suppression moderated by racial identity, and demonstrated a negative effect of descriptive norms in the context of that suppression. We speculate that this effect represents an outcome of psychological reactance (Rosenberg & Siegel, 2018; Steindl, Jonas, Sittenthaler, Traut-Mattausch, & Greenberg, 2015) among weakly racially identified AA men. For those with similar perceptions of descriptive norms, stronger perceptions of injunctive norms will be associated with stronger inverse associations between descriptive norms and behavior. In other words, this represents a manifestation of weakly identified AA men’s reactance to group-relevant descriptive norms as behavioral injunctions increase. A recent study attributed a similar negative effect of normative perceptions on cancer screening behaviors to psychological reactance (Sieverding et al., 2010). For our study, the items assessing norms referred to “most people who are important to me”, and did not mention anything about racial group (e.g., “Black people whose opinions matter to me…”). However, given the racially homogenous small groups in which the data were collected, and the presence of other items on the survey assessing race-related constructs, it is likely that race was salient when responding to survey items, thus priming the contexts for those men inclined to reactance. When AA men strongly identified with their racial group, no similar reactance was evident given a direct effect of descriptive norms on PSA testing behaviors. This is likely due to stronger motivations for group norm-consistent behaviors for individuals with stronger group identities (Crisp & Beck, 2005; Marques et al., 1988). Future research should examine whether the moderating effects of racial identity are equivalent for men and women using screening behaviors applicable to both.

Whereas most cancer screening behaviors are similar in that they may be influenced by perceptions of relevant constructs such as cancer fatalism, cancer risk perceptions, and knowledge (Manning, Purrington, Penner, Duric, & Albrecht, 2016; Manning, Wojda, et al., 2016; Mitchell, Manning, Shires, Chapman, & Burnett, 2014), they differ in other attributes. Variation in attributes of behaviors have been shown to moderate both the deliberative and non-deliberative effects of subjective norms and behaviors (Lapinski & Rimal, 2005; Manning, 2011a; Rimal, Lapinski, Turner, & Smith, 2011). Similarly, variation in attributes associated with cancer screening behaviors likely influence the associations between norms and behaviors. For example, for some individuals, salience of the discomfort associated with mammography and aversion to some CRC screening modalities may attenuate the effects of subjective norms on behavioral decision-making (Hawley et al., 2008; Kurtz, Given, Given, & Kurtz, 1993; Nadalin, Maher, Lessels, Chiarelli, & Kreiger, 2016; Watts, Vernon, Myers, & Tilley, 2003) more so than, for example, the relative ease associated with PSA tests. Variability in uncertainty given changes in screening guidelines, and uncertainty about the utility of screening to prevent cancer-related death, may also influence the extent to which individuals deliberatively or non-deliberatively use normative information for screening decision making (Manning, 2009). We currently lack a theoretically refined model of how variability among behaviors moderates normative influence; however, ongoing topical research (Manning, 2011a; Rimal et al., 2011) shows promise for such a model and its applications to health behaviors.

Interventions to increase cancer screening intentions and behaviors among AAs are important tools to address disparities in cancer incidence and mortality. Targeted interventions have been disseminated via multiple modalities such as patient navigators, peer-to-peer in community organizations, and print-based and computer-based media (Leone et al., 2013; Philip, DuHamel, & Jandorf, 2010; Rawl et al., 2012; Rogers, Goodson, Dietz, & Okuyemi, 2016; Sly, Edwards, Shelton, & Jandorf, 2013). Given the heightened role of subjective norms in cancer screening behavioral decision-making for AAs, normative messages could be effectively used in such targeted and tailored messages to promote cancer screening among AAs. This may be particularly effective in light of the unique influence of non-deliberative effects of norms among AAs demonstrated here, and given that norms have been shown to influence behaviors even when recipients of norm-based interventions did not perceive that they would (Nolan et al., 2008). In the area of cancer prevention, we only identified one study that used a social norms marketing approach to improve sun protective behaviors by influencing normative perceptions (Reid & Aiken, 2013). Given the relationship between screening behaviors and disparities in cancer incidence and mortality (The Center to Reduce Cancer Health Disparities, 2008), the potential that normative perceptions may serve as an efficacious focus for targeted intervention for AAs provides rationale and motivation for further investigations.

Interventions that target normative perceptions should consider two contemporary issues that may influence perceptions of screening behaviors among AAs. First, despite the historically lower screening rates for AAs compared to EAs, recent data suggest that cancer screening rates have converged for women. For example, data show that rates of breast and cervical cancer screening converged and were becoming slightly higher for AA women compared to EA women (American Cancer Society, 2017; Hewitt, Devesa, & Breen, 2004) and some evidence showed attenuations in racial differences in breast and colorectal cancer screening for women, but not for men (Rao, Breen, & Graubard, 2016). Second, some studies suggest that racial differences in cancer screening rates were largely due to sociodemographic variables such as income, education and health insurance status (Doubeni et al., 2010; Fisher et al., 2004; O’Malley, Forrest, Feng, & Mandelblatt, 2005). In some cases, AA cancer screening disparities were attenuated or even reversed when sociodemographic variables were accounted for (Rakowski, Clark, Rogers, & Weitzen, 2009; Rakowski, Clark, Rogers, & Weitzen, 2011). These data challenge some narratives regarding lower cancer screening rates among AAs, and such challenges may indirectly influence normative perceptions of screening behaviors. Studies that examine discrepancies between actual and perceived cancer screening rates for AAs could have considerably applicability and overlap with research examining how norm-based interventions may further improve cancer screening rates.

These data are limited by the fact that the studies were not explicitly designed to test hypotheses about normative perceptions and cancer screening behaviors among AAs. Nonetheless, these data presented an opportunity to examine our hypotheses given that they all had similarly operationalized constructs consistent with the TPB. Future research should be designed explicitly to further examine between-race differences in the influence of subjective norms in behavioral decision-making related to cancer screening behaviors, and to elucidate the mechanisms and moderators involved in the heightened effects of norms among AAs.