Do women and men talk differently?
When I give lectures about talk, one of the most frequent questions from audiences is, ‘Do men and women talk differently?’ Indeed, it is often not asked, simply asserted – ‘but, of course, men and women talk differently’. People who raise this topic generally assume I will confirm what they think they already know.
We think we already know that women and men talk differently for many reasons. We know it from the thousands of ‘pop’ psychology and communication books, from Men Are from Mars, Women Are from Venus to He’s Just Not That into You. There are thousands of scientific articles that report sex differences in behaviour of all kinds. We think we know about gender difference because research that does not show difference is often not published. Indeed, using gender as a research category is often not interrogated at all. That women and men talk differently has long since passed beyond the realm of academic enquiry and is just what we know.
Researchers have addressed questions about gender and language for several decades. In the 1970s, the linguist Robin Lakoff identified a so-called ‘women’s register’. It is through language, Lakoff argued, that women’s inferior place in society is maintained. Women’s talk is more polite than men’s; women use more tag questions (e.g., ‘isn’t it?’), use weaker directives, avoid swearing, and use more empty adjectives (e.g., ‘cute’) than men.
Lots of research followed Lakoff’s original ideas, testing out gender differences in who talks most, who interrupts more, who uses more ‘minimal responses’ (e.g., ‘mm’, ‘yeah’), and who controls the topic of a conversation. But many apparent findings about gender difference are built on shaky foundations.
For example, the idea that men grab power by interrupting women (more than women interrupt men) is based on a mistaken understanding of interruption. Consider Example 1. The square brackets on lines 02 and 03 indicate when the speakers talk at the same time.
Example 1: From Beattie (1983)
01 A: ... so he (.) he gives the impression that he
02 wasn’t able to train them up. [Now
03 B: [He didn’t try hard
04 enough heh heh heh
The original analyst of this example claimed that ‘Speaker B clearly interrupts Speaker A’. Another analyst, Celia Kitzinger, points out that, in fact, Speaker B starts talking at the precise point when Speaker A reaches the end of their sentence – in technical terms, the end of a ‘turn constructional unit’. Speakers frequently – and unproblematically – start to speak at the end of another speaker’s sentence. B does not interrupt A.
In Example 2, Tony and Marsha are talking about their son’s vandalised car. Kitzinger points out that Tony – a man – starts his turn before Marsha – a woman – completes hers.
Example 2: Tony and Marsha
01 Tony: That really makes me mad.
03 Marsha: .hhh Oh it’s disgusti[ng as a matter of ] fact.
04 Tony: [ P o o r J o e y. ]
Some researchers would use these three lines of talk as clear-cut evidence of men’s dominance over women. However, a closer look reveals that when Marsha describes the vandalism as ‘disgusting’, her turn has already delivered something that could be complete. Adding ‘as a matter of fact’ is not necessary. So Tony is not interrupting when he says, in overlap, ‘Poor Joey’. The parents talk in tandem to share the same feelings about what has happened to their son.
A little later in the same call, Tony and Marsha are discussing a problem for their son, who is travelling on a standby flight from Marsha’s home back to Tony’s. On lines 05 and 06, both speakers talk at the same time.
Example 3: Tony and Marsha
01 Marsha: What time did he get on the plane.
02 Tony: Uh::: (0.2) I: don’t know exactly I think it was
03 arou:nd three o’clock uh something of that sort.
05 Marsha: Oh: maybe he g[ot s’m-
06 Tony: [He took it at four. Gerda says.
This time, Tony starts talking at a point in Marsha’s turn where she is clearly not finished – she has not completed her turn in terms of its grammar, intonation (there is no full stop) or action (she has not finished her speculation about her son’s circumstances). Tony interrupts Marsha. But, as Kitzinger points out, Tony’s interruption is cooperative:
In response to Marsha’s enquiry about when their son boarded a plane, Tony reports not being sure and then gives an approximate answer (‘around three o’clock or something of that sort’). At line 05 Marsha is apparently using this report as the basis for some speculation about what may have happened and it is this speculation, based on his (as it turns out erroneous) report, that Tony interrupts. Tony’s new partner, Gerda, who has overheard his report of Joey’s departure time, has apparently corrected him and he is relaying this correction (‘four o’clock’) to Marsha before she can develop a theory about Joey’s flight based on incorrect information. The interruption is here used to implement a correction that has consequences for the turn Marsha is in the course of producing, and it is thus clearly a cooperative action, and from Marsha’s point of view, a helpful interruption.
All of this shows us that simply counting instances of overlapping talk says very little about interruption.
Other claims about gender differences in talk have also later been revealed to be flawed. A common assertion is that women talk more than men. The words that characterise the way women talk perpetuate this myth – women chatter, gossip, prattle, natter – about trivial matters. At the same time, feminist researchers have claimed to find that men talk more than women, hogging the floor in public and workplace settings. Yet more scientists have found that no gender difference exists in who talks more.
Throughout my book Talk: The Science of Conversation, there are hundreds of examples of men, women, salespersons, mediators, police officers, clients, customers . . . using tag questions (e.g., ‘don’t you’), ‘minimal responses’ (e.g., ‘mm’, ‘yeah’), initiating topics, and overlapping each other. If we categorise all of our examples in terms of gender alone, we will make erroneous claims about women and men at the same time as we miss what is going on in each case. We will stop looking for other explanations for the way people talk.
Consider two examples from the start of a telephone call to a double-glazing company (Example 4) and the vet’s (Example 5).
Example 4: Calling double-glazing sales
01 Sales: G’d afternoon, Fine Bar Wi:ndows,
03 Customer: .shih (.) .hh >hi< w’d it be po:ssible f’somebody
04 t’come an’ give me a quote on uh: a window an:
05 some doors please.
Example 5: Calling the vet
01 Vet: Dunnetts Vets.=Highuptown, Maggie speaking, how
02 can I ↑he:lp.
04 Pet owner: Hello there:.=um: I need t’make an appointment
05 t’bring the cat in t’get its um: updated ↓vaccina:tions.
In both cases, the callers are making a request. They design their requests differently. For some gender difference researchers, the customer calling about windows uses stereotypically female language, with hedges and modal verbs (‘Would it be possible’), hesitations (‘.shih (.) .hh’), and in a polite manner (‘please’). The pet owner is stereotypically male, with a direct request (‘I need . . .’) which is less polite.
The caller in Example 4 is a man; in Example 5 a woman. Perhaps these are just atypical cases – obviously not all men and women speak like their stereotype! But if we start and stop with gender as the only lens through which to interpret the data, we identify only exceptions that prove the rule.
Instead, conversation analysts start with action – what are the speakers doing? Are they entitled to do it? Do they know about the service and what it provides? What is at stake? People begin requests with ‘I need’ when, say, calling for an ambulance, or a doctor, or
for something relatively important and urgent – like vaccinations. People also say ‘I need’ when they know what the service offers, and that they are entitled to use it. However, when people are less familiar with the service and what it offers, or they are less entitled to ask for it, or if their problem is not urgent, they often ask ‘would it be possible’. These kinds of contingencies better explain the differences in the way people talk.
Plenty of myth-busting science about gender difference exists. Yet our ideas about differences between women and men persist. Gender difference studies start with the assumption that all women and all men can be classified, even with caveats, into two homogeneous groups. Often, then, researchers do not really study gender difference. They simply create and maintain it. They make selective observations to confirm what we already know about how women and men behave.
As a PhD student in the early 1990s, I struggled with all these ideas. I wanted to say something about gender and language. But I didn’t want to make assumptions about what I would find when I analysed interaction. I didn’t want to reproduce stereotypes. At the same time, I felt that gender imbalance and stereotypes were all around me. How could I capture them? The conversation analyst Curtis LeBaron had the answer:
we should not . . . say ‘oh, look, here’s a man and a woman talking; . . . oh, we can make these conclusions about gendered communication’. But rather we should say, ‘gender only becomes an issue when the participants themselves make it one and we can point to different things about that’.
And, it turns out, people invoke gender a lot. Gender matters to people. Making it relevant does things. Here is just one example, from the PhD I eventually wrote. My research was on gender and conversation in university tutorials. My initial interest was in whether or not male students dominated the floor, leaving female students less verbal space to participate.
My data were thirty hours of tutorials. I quickly realised that coding and counting and correlating with gender would tell me little about what was going on in the encounters. I was also increasingly unhappy with the Men Are from Mars . . . view of gender. So I asked LeBaron’s question: how does gender become an issue for participants themselves?
In Example 6, four psychology students are analysing an image in a magazine. They have been instructed to produce a collaborative written report about the materials they are discussing. Bob reminds them of this task.
Example 6: University tutorial
01 Bob: Is somebody scribing. who’s writin’ it.=
02 Ned: =Oh yhe:ah.
04 Matt: Well you can’t [ read my ] writin’ once I’ve
05 Ned: [She wants to do it.]
06 Matt: [wri:tten it.]
07 Klara: [.hehhhh ]
08 Ned: We:ll secretary an’ female.
As the person asking the question ‘Is somebody scribing. who’s writin’ it’ – taking the ‘first pair part’ – Bob relieves himself of the obligation of answering the question and becoming the scribe. Ned is first to respond, saying, ‘Oh yhe:ah.’. He remembers that they must do this, but does not volunteer. After a gap, Matt provides a reason why he cannot be the scribe (his writing is indecipherable). At the same time, Ned nominates Klara. He says, ‘She wants to do it.’ as he points to her. Ned’s words and embodied conduct are fitted together.
Klara has also not volunteered to scribe. Her small laughter particle (‘.hehhhh’) could be responding to Matt’s comment about his own handwriting, or Ned’s suggestion that she wants to scribe. Ned then provides an account for his nomination of Klara: ‘We:ll secretary an’ female.’. At this point, gender becomes relevant to the encounter. It is invoked by Ned to do something. The stereotype that women are secretaries is used at a particular point to make something happen. And Klara becomes the secretary. Her participation in the discussion is reduced to note taker.
You might agree that this is a powerful moment in which gender and language come together. You might also think that this is a oneoff. So what? It is much easier to divide the world into two genders from the start and count the things people do. How do we capture sufficient numbers of moments like this to say something meaningful and generalisable?
There are several answers to this question. First, Example 6 is a demonstration of a method of investigating gender and language – one that does not start with any assumptions about women and men. We can capture moments of gender relevance without using gender as a scientists’ category, at all. We can investigate gender as a speaker’s category.
Second, conversation analysis is neither a quantitative nor a qualitative method. It is sometimes both. One conversation is not a useful unit of analysis. One conversation contains many questions, many answers, many overlaps, many pauses, many ‘ums’ and ‘uhs’. An apparently single conversation delivers lots of instances of particular phenomena. In the conversation analyst Emanuel Schegloff’s words, ‘one is also a number, the single case is also a quantity, and statistical significance is but one form of significance’. Claims about gender and interruption may be based on large datasets, but if there is a basic error in identifying interruption in the first place, Schegloff is correct to point out that ‘quantification is no substitute for analysis’.
Third, conversation analysis is a form of logical analysis. An example from linguistics makes the point: the sentence ‘Peter forwarded the letter to his aunt Mary’ is grammatical (if you don’t agree with me, you don’t understand what ‘grammatical’ means, or you don’t speak English), and its grammar can be sensibly analysed. On the other hand, ‘To aunt the forwarded Peter letter Mary his’ contains the same words but is not grammatical, and has no structure. It is not a matter of counting how many times people say these things, or asking how many people agree about their grammaticality. It’s a matter of knowing how to speak English, and you can make definitive analyses of these things from just one example.
What makes a turn in conversation analysable is that it is recognisable and understandable by people, including analysts, who are members of a culture and a linguistic community that talk in those ways. We don’t need huge samples and probabilistic statistics to do an analysis, even though we do need collections of instances of a phenomenon to analyse and figure out how they work. But even then, the analysis of each and every ‘instance’ is done on the same basis, including recognising something as an instance, which is the tacit ability that people have, and that anthropologists need, to understand the uses of their own natural language, or one in which they have acquired some competence.
Fourth, when people ask me a ‘how many’ question, I respond with an illustration to challenge the presuppositions built into it. I ask how many black holes (or big bangs) does a physicist need to say something meaningful about the science of black holes? Most people laugh, and then say ‘one’. I recently asked a physicist this question. He looked thoughtful for a moment, and then replied ‘five’. So how many negotiations with persons in crisis do we need to study before saying something meaningful about effective practice? How many times would you need to see that asking people to talk does not get people to talk? I do have more than one example. And it is helpful to show that asking people to talk does not work in other settings too. But the logic and explanation remain the same.
Fifth, quite simply, it would be strange to discount an analysis of a conversation because there is only one instance. Sometimes there may only be one instance to study. If you are at a party, and a guest collapses, hopefully you or someone else will call for an ambulance. If the ambulance takes a long time to arrive, and the guest dies, the telephone call will be a source of evidence in any subsequent investigation. Compared to the numbers of patients calling their GP receptionist, there are very few police negotiations with suicidal persons in crisis. But it is important to know how they work.
Finally, I have spent a good chunk of my career showing that, actually, one can identify and analyse moments like ‘Well secretary an’ female’ systematically. With large datasets of lots of interaction, it turns out that people often invoke gender (or other categories such as age or ethnicity) in the same kinds of settings, in the same kinds of turns, doing the same kinds of actions. We just need to study real talk – the answers are often right in front of us.
Elizabeth Stokoe is Professor of Social Interaction in the Discourse and Rhetoric Group at Loughborough University, using conversation analysis to understand how talk works. Outside the University, she runs workshops with professionals using her research-based communication training method called the ‘Conversation Analytic Role-play Method’. She is one of thirteen WIRED 2015 Innovation Fellows; has given TEDx, New Scientist, SciFoo/Google, Cheltenham Science Festival and Royal Institution lectures, and her research and biography were featured on BBC Radio 4's The Life Scientific. Her book Talk – The Science of Conversation, published by Robinson, is available now. This extract is reproduced with kind permission.
You can read more from Professor Stokoe in her Latitude Festival appearance for us (also available as a podcast); or revisit the 2013 'Careers' interview which she says led to all sorts of new opportunities!
Professor Stokoe also introduces a special issue in our December 2018 edition.
See also the recent profile in The Observer.
BPS Members can discuss this article
Already a member? Or Create an account
Not a member? Find out about becoming a member or subscriber