Offenses and apologies

« previous post | next post »

In last week's news, there was a fair amount of interest in a study finding that women apologize more than men do. Curiously, there has been no coverage so far, as far as I can tell, by the New York Times, by the Washington Post, by BBC News, or by NPR. I hope that this is not because the reporters, editors, and pundits at these more serious publications collectively decided, contrary to their usual judgment in such cases,  that this was a bit of gender-stereotyping fluff better left to the tabloids.

In my opinion the study was well designed and well done (with the usual caveat that the subjects were all 20-ish undergraduate psychology students), and the paper documenting it was clear, careful, and thought-provoking.  Also, for a change, the press coverage seems to have been fairly accurate so far. I suspect that this is a tribute to the paper's lucidity, and perhaps also to the communication skills of Karina Schumann, who was the first author and the person most often interviewed by reporters.

The paper was Karina Schumann and Michael Ross, "Why Women Apologize More Than Men: Gender Differences in Thresholds for Perceiving Offensive Behavior", Psychological Science, published online 9/30/2010. It documents two experiments, of which the first was a diary study involving 33 female and 33 male students from the University of Waterloo:

Participants were asked to complete two sections of an online questionnaire every evening for 12 consecutive nights. For the transgressor section, participants described up to three instances that day in which "you apologized to someone or did something to someone else that might have deserved an apology (regardless of whether or not you apologized)." For the victim section, participants described up to three instances that day in which "someone else apologized to you or did something to you that might have deserved an apology (regardless of whether or not he or she apologized)." The order of the transgressor and victim sections was counterbalanced across participants. If participants could not recall an event for the first section, they proceeded to the second section. If they were unable to remember an event for the second section, they terminated the session.

For each event, participants reported what occurred, who was involved (including gender information), and whether they had apologized or received an apology. Participants reported the exact wording of the apologies to the best of their abilities.

The women reported giving 37% more apologies than the men did (217 vs. 158). They also reported committing 36% more apology-worthy offenses than the men did (267 vs. 196), so that the men and women apologized for the same proportion of (what they reported to be) apology-worthy offenses (81.3% vs. 80.6%).

In the other direction, the women reported getting 28% more apologies than the men did (142 vs. 111), and being the victim of 48% more offenses than the men (242 vs. 164).

The paper's basic conclusion:

The diary data suggest that women offer more apologies than men do because women have a lower threshold for what constitutes offensive behavior. If so, gender differences in the perceived severity of offenses may lead men and women to have different views about whether or not an apology is warranted. As both transgressors and victims, women are more likely than men to judge offenses as meriting an apology.

The most interesting part of this study, for me, was the way that the researchers coded the reported offenses and the reported apologies. For the offenses,

[T]wo independent coders categorized the offenses into four types, adapted from the offense types described by Holmes (1989): (a) relational (e.g., insulting someone), (b) failed obligation (e.g., failing to complete chores), (c) inconvenience (e.g., calling a wrong number), and (d) physical or material (e.g., bumping into someone, damaging someone's belongings). Interobserver reliability was high (κ = .94).

It's impressive that they coded everything twice and calculated the kappa statistic, which is a useful measure of inter-rater reliability. It's also impressive that their kappa was so high.

On the apology side, they present an interesting featural decomposition:

A comprehensive apology contains as many as eight distinguishable elements: remorse, acceptance of responsibility, admission of wrongdoing, acknowledgment of harm, promise to behave better, request for forgiveness, offer of repair, and explanation. […]

Two independent coders (one male and one female) who were blind to participants' gender coded the apologies for the presence of each of the eight apology elements (κ = .92). Discrepancies between coders were resolved through discussion. We summed the number of elements in each apology to represent its comprehensiveness.

Again, the gender-blind double coding, with a high value for the kappa statistic, is impressive. And having coded the "comprehensiveness" of apologies, they were able to draw an additional conclusion:

[T]here was no gender difference in how men and women apologized. It appears that once men and women categorized a behavior as offensive, they were equally likely to apologize for it, and their apologies were similarly effusive.

The study's other negative results were also interesting, among them these:

Women and men did not differ in the proportion of offenses they reported as transgressors versus victims […] The proportion of each offense type reported did not differ as a function of gender or perspective […] Transgressor's gender did not significantly interact with the gender of the victim to affect the number of apologies reported or the proportion of offenses for which participants apologized.

There was one other positive (in terms of sex difference) result:

Men and women did not differ in the proportion of offenses they reported in the various relationship categories, except for offenses occurring between romantic partners. Women reported more [i.e. a higher proportion of] offenses occurring between them and a romantic partner (13.21%) than men reported (4.24%) …. Participants did not differ in the number of offenses they reported from the transgressor and victim perspectives within each relationship type.

Is this news? Well, yes:

According to various academic and popular writers, women apologize readily for their transgressions, whereas men do not. Some commentators suggest that women are too apologetic, but most presume that men are insufficiently contrite (Engel, 2001; Lazare, 2004; Tannen, 1996, 2001). These commentators offer a set of related explanations for the gender difference, the basic tenet being that men associate apologies with weakness. For example, Engel (2001) argued that men refuse to apologize because they have difficulty admitting they are wrong. She suggested that for men, admitting wrongdoing is like "losing a power struggle," and apologizing therefore hurts men's "delicate egos" (p. 49). Adopting a different argument, Tannen (1996) reasoned that women readily apologize because they are more concerned than men are with showing courtesy to other people. Whatever the interpretation, the bottom line is exemplified by the title of comedian Jim Belushi's (2006) book on manhood: Real Men Don't Apologize.

According to Schumann & Ross's (I think sensible and careful) interpretation of their experimental results, this is all wrong. In their data, at least, there's no sex difference in willingness to see oneself as being in the wrong, or in willingness to admit error, or in propensity to apologize given a perceived offense. Rather, (Waterloo undergraduate) women are somewhat more likely than (Waterloo undergraduate) men to perceive a given piece of behavior as offensive, and also are likely to perceive the degree of offense as somewhat greater, whether they're the transgressor or the victim. As a result, the women in their sample were somewhat more likely to see a given behavior — their own or someone else's — as a transgression requiring apology.

Schumann and Ross speculate sensibly about the reasons for this difference:

What is the psychological basis of gender differences in perceptions of the severity and frequency of offenses? One possibility is that women might perceive more offenses because they are more focused on the experiences of other people and on maintaining harmony in their relationships […] A second possibility is that men have a higher threshold for both physical and social pain.

This sort of study is on the boundary between linguistics and social psychology — and it would be nice to see more interdisciplinary communication and collaboration across that boundary.

[I should note the usual problem with potential misinterpretation of generic plurals. One thing that's lacking in this paper is a clear picture of how the within-sex variation compares to the differences in male and female mean values.]



22 Comments

  1. groki said,

    October 10, 2010 @ 7:40 am

    glad to hear that the study was so well done, and that the results support a good deal of gender parity.

    still, it seems to me there is a non-trivial connection between, say, for men, admitting wrongdoing is like "losing a power struggle" and men have a higher threshold for […] social pain.

    maybe Real Men Feel They Can't Afford Too Much Need To Apologize.

  2. xah lee said,

    October 10, 2010 @ 7:59 am

    nice article. Good to know that some study on the sexes are done well.

    though, it's not too clear to me how's linguistics come into play here?

  3. Mark Mandel said,

    October 10, 2010 @ 9:14 am

    xah lee @ 7:59 am: It's a study of the use of language.

    I just apologized to our cat. I heard some loud clattering behind me in the kitchen and said, "All right, cat, what have you done now?" I turned and saw that she was hunkered down on the living room carpet in her usual cool-weather position, and realized that what I'd heard was the ice-maker in the freezer. "I'm sorry, cat, that wasn't you at all."

  4. Jon said,

    October 10, 2010 @ 9:57 am

    @groki: "maybe Real Men Feel They Can't Afford Too Much Need To Apologize."
    That ignores the finding that it is in both the giving and receiving of offence that men were found to be less sensitive.

    Mark said: "One thing that's lacking in this paper is a clear picture of how the within-sex variation compares to the differences in male and female mean values."
    Absolutely! Whenever I read about surveys discussing differences between men and women, I think of the graph of men's and women's heights, with a large overlap between the two distributions, to remind myself that real differences in the average of a characteristic may tell you nothing about individuals.

    Having said that, I do often think that my wife apologises too much, and takes offence too readily. Maybe we're both just average.

  5. marie-lucie said,

    October 10, 2010 @ 10:02 am

    This study was done at the University of Waterloo, which is in Ontario (Canada). Canadians (at least anglophone Canadians) are supposed to be more apologetic than Americans.

  6. Henning Makholm said,

    October 10, 2010 @ 11:31 am

    The study concludes that women are more likely than men to find any given interaction apology-worthy. It seems to me that the observations this is based on could be equally explained by the hypothesis that the female subjects either participated/observed in more instances of social interaction during the day than men, or remembered more of them when time came to fill out the questionnaire, but that the two genders had statistically similar reactions to any [i]given[/i] situation.

  7. Kutsuwamushi said,

    October 10, 2010 @ 11:39 am

    xah lee: Sociolinguists do a lot of work like this. Studying how society influences what we say and how we say it is one of their main purposes in life their research. There's often a lot of collaboration/overlap with psychology.

    What puzzles me about this is how it's being treated as news. I thought that it was commonly accepted within linguistics that women apologize more than men (at least in the Western Anglophone countries). Did I mistake a stereotype for something backed up with real research until now?

    [(myl) As the paper explains, the idea that women apologize more has often been asserted but never (I believe) verified empirically — and a high proportion of such stereotyped beliefs turn out to be false when tested. But the more interesting and important part of this work, as explained in the post, is the evidence about the reasons for the fact — what they are and what they aren't.]

  8. Robert Coren said,

    October 10, 2010 @ 12:16 pm

    @Mark Mandel: Ah, but your cat is female, and from what I've heard she has a very low threshold for what's considered "offensive", at least when she's on the receiving end.

  9. Fred said,

    October 10, 2010 @ 4:16 pm

    "One thing that's lacking in this paper is a clear picture of how the within-sex variation compares to the differences in male and female mean values."

    Is the study large enough for such a comparison to be meaningful? (I never was very good at statistics).

    [(myl) With 33 members of each sex, it should be possible to get a decent estimate of the variance as well as the mean (e.g. of the number of offenses and apologies reported by males and by females), which would allow an effect size to be calculated. And a simple scatterplot (e.g. of the relation of those two quantities) would be visually informative.]

  10. Ned Danison said,

    October 10, 2010 @ 4:52 pm

    1. I agree with Henning Makholm. Maybe women just interact more and have more opportunities to apologize.

    2. Anecdotally, in Chinese and American cultures, the two cultures I am familiar with, women seem generally more concerned than men with finer points of social propriety. As mothers, women seem to be more involved in religious activities and more concerned than fathers with instilling a more detailed sense of appropriateness in children (e.g., writing thank-you cards, etc.). Women seem more concerned with social "oughts". With a more finely-tuned sense of the way things ought to be, maybe women think there ought to be more apologies.

  11. Rubrick said,

    October 10, 2010 @ 7:03 pm

    It is nice to hear of such a study being done relatively well — though I find it disappointing that it's based at root on self-reporting. Unless I'm misunderstanding the design, a greater ability among (say) women to recall apology-related incidents that took place earlier in the day could strongly bias the results.

    [Looking now at the paper's abstract, it seems that Study 2 should serve as a hedge against this possibility. Nonetheless, it's a pity that these things can't use as their starting point a recording of the entire day's actual interactions. I realize there are practical and ethical difficulities with this.]

  12. Igor Yanovich said,

    October 10, 2010 @ 9:21 pm

    "According to Schumann & Ross's (I think sensible and careful) interpretation of their experimental results, this is all wrong." (emphasis mine)

    I'm afraid this is a way too strong. Here is a simple model in which both the data from the paper are predicted, and it is true that "men associate apologies with weakness".

    1. Men associate apologies with weakness (more than women do – add the same to each subsequent clause, too.)
    2. Men avoid putting either themselves or the people around them into situations where people would feel weak.
    3. Encountering a potentially offensive situation, men interpret it trying to minimize the risk of anybody's feeling weak.

    Of course, it is possible the actual world is not an instantiation of this simple-minded model, but the model is not ruled out. I do not know (I have not looked into the paper itself) if the phrase I cited was your interpretation or essentially their interpretation, but it is not right.

    I guess a better characterization of the results would be something like that: it is not that men judge their own potentially offensive behavior more leniently than others' potentially offensive behavior (in the self-reporting setting). What is ruled out is not a potential reason why men apologize less, but a hypothetical possibility that in a number of situations they do not want to apologize themselves, but expect apologies from others (again, as self-reported.)

    [(myl) You've misinterpreted the passage I've quoted. The explanation that Schumann and Ross attribute to Engel, for example, clearly implies that men will tend to apologize for a lower proportion of (what they perceive as) their offenses than women do — which is contrary to what Schumann and Ross found.

    In their second experiment, Schumann and Ross test their threshold hypothesis more directly, by asking subjects to evaluate the seriousness and apology-worthiness of both hypothetical and recalled offenses. You may continue to insist that an association of apologies with weakness, and a desire to minimize weakness not only in themselves but in others, is the reason that the male subjects in this experiment might choose to see offenses (whether by themselves or others) as (on average) less serious and less apology-worthy than the female subjects. This is indeed a possible explanation, though not a very plausible one.]

  13. Ben Hemmens said,

    October 11, 2010 @ 5:58 am

    Maybe the men were just pretending to be tougher. The difference in rate of reported "offense" events is the stuff they were embarrassed to even write down.

    Or maybe they were just lazier about diary-keeping in general.

    Or have trouble remembering what they did during the day. I mean, I have bought task-managing software so I can tick off things that I do, partly so I can tell my wife what I've been up to while she was away ;-)

    Try giving the men an iPod with an app for logging the incidents closer to real time.

    [(myl) The app idea is worth trying, though (believe it or not) only a moderate fraction of undergraduate psych majors have iPhones, and if you limit participation to those who do, you risk getting a biased sample.

    The "men are lazier" (or maybe less diligent) idea occurred to me, and also to the researchers, who checked it by comparing the total number of diary entries (out of 12 possible):

    Men (M = 9.78, SD = 2.29) and women (M = 9.14, SD = 2.65) completed a similar number of entries, t(64) = 1.08, p = .28.

    If I understand what you're suggesting, the "pretending to be tougher" idea doesn't work, at least in a simple form: recall that men and women reported apologizing for exactly the same proportion of recalled offenses. So you'd have to assume that "pretending to be tougher" entails failing to write down offenses and apologies in the same proportion. I agree, though, that it would be nice to support diary studies with studies that involve direct observation of behavior, and also to try to differentiate pretending to have a higher social-pain threshold from actually having a higher threshold, for instance by using appropriate physiological measures.

    The "males have worse episodic memory" idea would work, and is even supported to some extent by e.g. Herlitz et al., "Gender differences in episodic memory", Mem Cognit 25(6):801-11, 1997 (.pdf). However, in Herlitz et al.'s test of free recall of recent activities, the sex difference was too small to explain Schumann and Ross's basic finding. Out of 31 possible activities, women on average recalled 8.67 while men recalled 8.04 (effect size d=0.19), which is a much smaller proportional difference (8%) than Schumann and Ross found (36-37%). But obviously, the experimental conditions were quite different.]

  14. Ben Hemmens said,

    October 11, 2010 @ 10:03 am

    "The app idea is worth trying, though (believe it or not) only a moderate fraction of undergraduate psych majors have iPhones"

    That's why I said give them the gadget.

    "So you'd have to assume that "pretending to be tougher" entails failing to write down offenses and apologies in the same proportion."

    I don't think so. When they sit down in the evening to write their diary, they just do less soul-searching about what might have been a relevant incident. If they were prodded to think over their day more deeply, they might well come up with more. But I think there could be a male tendency to just shrug things off, one way or the other.

    Did the study do anything about controlling the rates of social contacts per day each group had? Women tend to dot around doing errands (well known from urban mobility research: women and children visit many more destinations for more activities in the course of a day than do men, on average). This might apply to students, too. What proportion of the women and the men were at college near where they grew up (may be involved in interactions with family & old neighbourhood) or far away from home (involved in college and social life consisting of seeing just a few friends).

  15. Katie said,

    October 11, 2010 @ 10:35 am

    One thing that I think would be interesting that might not come up as strongly in a study of college students is apologies in workplace or hierarchical interactions. Where I most often hear the "women apologize more than men" assumption voiced is when people are saying that women have a tendency to apologize for every thing that goes wrong at work while men don't (see, e.g., that Citibank list). This generally seems to be interpreted as a sign of submissive behavior in such situations. I wonder if you'd see a greater difference in apology-levels when women vs. men are interaction with bosses and teachers than you do with peers and romantic partners. If you didn't or if you did and, as in this study, the difference was still linked to perception of behavior as offensive rather than how men and women are personally attempting to position themselves, that would go against conventional wisdom, I think.

  16. Mary Dee said,

    October 11, 2010 @ 4:00 pm

    I wonder if there is some generational differences that are being overlooked by this study. Most of the earlier reports of women being more apologetic would have used men and women of a different generation from today's undergraduate psychology majors. There are other differences in expectations and behavior comparing today's undergraduate men and women to those of several decades ago.

  17. Kaviani said,

    October 11, 2010 @ 5:52 pm

    A comprehensive apology contains as many as eight distinguishable elements: remorse, acceptance of responsibility, admission of wrongdoing, acknowledgment of harm, promise to behave better, request for forgiveness, offer of repair, and explanation.

    Is that a Canadian thing? I only see an apology as an expression of remorse or regret. I'm also a male, so that could play into the interpretation as well. (I also tend to be very suspicious of gratuitous, kneejerk apologies. I certainly do not give them.)

    Per personal observations, I figure women over time had to apologize more or get physically attacked in eons past. Presently, it seems like a part of being liked, which most females seem to associate with "being nice".

  18. Igor Yanovich said,

    October 12, 2010 @ 1:57 am

    [(myl) You've misinterpreted the passage I've quoted. The explanation that Schumann and Ross attribute to Engel, for example, clearly implies that men will tend to apologize for a lower proportion of (what they perceive as) their offenses than women do — which is contrary to what Schumann and Ross found.]

    Thanks for pointing that out. After actually looking into the article, I see that S&R's specific alternative hypothesis for the first experiment was that "men apologize less often because they are unwilling to admit wrongdoing (Engel, 2001)". This hypothesis, of course, is as solidly ruled out as possible given the design.

    But it hardly counts as the proof that "this is all wrong", and S&R themselves did not actually construe their argument that way, so I think I was partly right, even having misunterpreted the passage. S&R show that perceived level of offense correlates with the apology-worthiness, but only speculate (on p.6) as to why there seem to be different threshholds for women and men.

    [(myl) In their second experiment, Schumann and Ross test their threshold hypothesis more directly, by asking subjects to evaluate the seriousness and apology-worthiness of both hypothetical and recalled offenses. You may continue to insist that an association of apologies with weakness, and a desire to minimize weakness not only in themselves but in others, is the reason that the male subjects in this experiment might choose to see offenses (whether by themselves or others) as (on average) less serious and less apology-worthy than the female subjects. This is indeed a possible explanation, though not a very plausible one.]

    Well, _if_ there is an association of apologies with weakness, it is likely to exist at the level of Bourdieu-style bodies. It would then be expected that both hypothetical and recalled offenses are perceived similarly. As for whether the weakness explanation is a plausible one, I'd rather not pass a verdict one way or the other: there is no hard evidence either way that I know of, and intuitive plausibility judgments are dangerous to rely on when they concern a strongly stereotyped area.

  19. ronmurp said,

    October 12, 2010 @ 6:21 am

    Doesn't this need a different study that doesn't rely on reported incidents?

    I'd have thought contrived experimental Male-on-Male, M-F, F-F interactions, with opportunities cause offence, take offence and apologize, would be more revealing.

    There just seem to be too many external world variables in self-reporting like this.

  20. Mark F. said,

    October 13, 2010 @ 12:32 pm

    Ben Hemmens — You write "But I think there could be a male tendency to just shrug things off, one way or the other." Isn't that what the study purports to demonstrate?

  21. John said,

    October 13, 2010 @ 5:14 pm

    >>(with the usual caveat that the subjects were all 20-ish undergraduate psychology students)

    I don't have access to the article, but does it specify that the subjects were psych students? I studied math there a few years ago (and still live in town), and psychology studies were generally open to students in any program.

    [(myl) The subjects for the first experiment came from what is described as "the University of Waterloo psychology department's participant pool", and were paid for their participation. At my institution, and some others that I'm familiar with, the psych department's participant pool is mainly recruited from psych department courses (though I also recruit in Linguistics 001, so that my fellow linguistics department members can access the pool for subjects). In most cases, independent of being paid, students are given some sort of course credit — for example, a certain number of hours of service as a subject may be a course requirement.

    For their second experiment, the researchers specify that "Sixty-three female and 57 male undergraduates (ages 17–27 years, M = 19.62, SD = 1.98) participated in exchange for course credit." I rather doubt that any engineering courses gave course credit for participation in such studies, though I could be wrong. Of course, engineering students may well have taken psychology courses.]

    The University of Waterloo (UW to us Canadians) is a respected math/computing/engineering school and has quite competitive programs in these fields, which happen to draw more males than females. I think it's also reasonable to claim that the people in high-end technical programs have a higher chance of being described as 'socially awkward or oblivious'. My point being that, speaking in sweeping generalities, a male at this school is less likely to pick up on any kind of subjective social event than is a female at this school. Not just whether an event deserves an apology.

    [(myl) Good point. See here for some data on the distribution of Simon Baron-Cohen's "systematizing quotient" and "empathizing quotient" across various academic disciplines.

    In some sense, it's not possible to avoid bias of this sort, since no matter what group you take your subjects from, it will have a distribution of characteristics that may be very different from that found in some other groups.]

    I guess I believe the findings of the paper intuitively for the general case, but I think the UW population (which I expect was sampled from) has some bias built into it which I bet would make the results more pronounced than they would be in other populations. If I'm wrong, well done to the researchers anticipating this bias! :)

    There is a saying on campus among the single women at UW, about finding a man at this male-heavy school: The odds are good, but the goods are odd.

  22. maidhc said,

    October 14, 2010 @ 4:44 am

    Here's another one, not supported by structured research yet. In an environment where language learners mingle with native speakers, language learners apologize more than native speakers.

    Make the environment a crowded bar where one is trying to make one's way to order a drink.

RSS feed for comments on this post