This post was a response to a decent (though not too exciting) study and the horrible media reporting on it. As the blogosphere focused on the press releases, I decided to look at the paper itself and see what it really says. It was first posted on August 09, 2005. Under the fold…
I saw this on Pandagon first – a response to an article on NeuroImage about gender-specific voice recognition. Actually, it was not a response to the article itself (behind the subscription wall), but to the MSM reporting about the article. Soon, other bloggers chimed in, notably Feministing, Blondesense, Lindsay and Amanda again.
Now, when you see the range of quality of reporting about this article, you will not be surprised. Here are some examples: Why ‘imaginary voices’ are male, Male and Female Voices Affect Brain Differently and It’s official! Listening to women pays off. Look at this one for a taste: Can’t hear you, dear … blame my brain:
WHETHER it is to do the dishes, clean the car or vacuum the living room, men now have an answer to their wife’s war cry that they never listen: it’s not me, darling, it’s my brain.
Scientists now have discovered that women’s voices are more difficult for men to listen to, and process information from, than the voices of other men.
I can just imagine some young guy fresh out of J-school, officially named the “science reporter” for the local paper because he actually took BIO101, playing X-box in his cubicle when the editor walks in and gives him a deadline. The poor guy checks his science news subscriptions and finds nothing interesting. But he has to write something…anything! And it better have an eye-catching title. What’s the best bet? Get something about the relations between the sexes – that is bound to be a good material to turn into a Homer Simpson joke. So he scans the articles until he finds one that mentions “sex” or “gender” and gets to work to mold it for the tastes of the mysognist proles.
This week, there was not much about gender, but he found one article – the one we are talking about:
Sokhia DS, Hunter MD, Wilkinson ID, Woodruff PWR. Male and female voices activate distinct regions in the male brain.
Who cares that it has NOTHING to do with gender or with relations between men and women. He can MAKE it be so. Of course, I wanted to see the actual article. Well, after I read it I can say it is pure neuroscience motivated by application for understanding schizophrenia. What did they do and how did they interpret their results?
In schizophrenia, patients often “hear voices” (or, in scientific parlance “auditory verbal hallucinations”). It has been established before that the voices are perceived as male 71% of the time and as female 23% of the time irrespective of the sex of the patient. The voices are also commonly perceived as coming from a middle-aged person, positioned to the right of the patient, derogatory in content, and possessing “BBC newsreader” accent in quality (in the UK, of course – I doubt this holds in China).
It is also well known that there are clear distinctions between male and female voices that are not solely based on pitch:
In humans, diverse speech parameters help define a wealth of
speaker-related attributes such as age, gender and emotion. At both
the phoneme level and sentence level of speech, the most salient
extra-linguistic acoustic parameters simultaneously used to perceive
gender from heard speech during normal human audition are:
(i) The vocal cord fundamental frequency, F0, perceived as
(ii) Formant frequencies, especially first and second formant
frequencies, F1 and F2 respectively; and
(iii) ”Other” parameters–such as spectral tilt, spectral density,
”aspiration noise”, frequency bandwidths and amplitude
difference between frequencies (Klatt and Klatt, 1990;
Titze, 1989; Wu and Childers, 1991).
Female voices generally have a higher pitch (i) as well as “harmonics” (ii) than male voices. However, there is an overlap in pitch that encompasses the highest tenors and the deepest contra-altos. It is also known that there is a difference between sexes in other parameters (iii), including prosody (intonation).
Woodruff’s group wanted to know why both men and women with schizophrenia predominantly identify “voices” as male. In order to see what areas of the brain are utilized during hallucinations, they first needed to identify parts of the brain that are normally used for identification of gender in speech. This is that first study.
They did two experiments, one purely behavioral, the second a brain-scanning experiment with fMRI. In the behavioral experiment, 33 males and 33 females were outfitted with headphones and given recordings of female or male voices uttering emotion-neutral sentences. The subjects were instructed to press one of the two buttons to indicate if they thought the voice was male or female.
Some of the male voices and some of the female voices were modified. They were placed in the gender-ambiguous range of pitch (the castrati-tenor/contra-alto range). As expected, when hearing gender-ambiguous voices both men and women took more time to make the male/female decision and made more mistakes. Still, the accuracy was still pretty high (except in cases in which a female voice was presented in male-only pitch or vice versa – there accuracy fell to as low as 20%). This suggests that previous research on the importance of “other parameters” (iii) is correct. Intonation (prosody) is quite sex-specific and can be used as trustwothy information in cases when pitch alone is insufficient (e.g., within the gender-ambiguous range of pitches).
In the second experiment they used the SAME recordings (both unmodified male and female and modified male and female) to monitor brain activity during auditory gender recognition.
All of the brain is active all of the time. All brain cells continuously recieve blood from circulation. However, an area of the brain that is more active (presumably the one that is involved in the task at hand) is thought to have an increased blood supply (some more capillaries are open). fMRI measures the blood flow in the brain.
The differences between less active and more active areas of the brain are miniscule. fMRI is usually used to compare two brain-states: “Stimulus” and “Non-stimulus”. In other words, the baseline blood flow through the brain measured while the subject is resting is subtracted from the blood flow through the brain measured while the subject is involved in an activity. The brain area in which such subtraction results in a number greater than zero (or statistically equivalent to zero) is pronounced “active”.
This type of research is very difficult to do, not to mention extremely expensive. All sorts of things go through the minds of subjects. The brain activity in various parts of the brain fluctuates all the time. There is great inter-individual variabiltiy in the way the brain is built during the development and subsequently used. Thus, it is very difficult to get clear-cut data. A large sample-size is needed, but is often limited by constraints of money and time.
The sample of 12 in this study is actually pretty big for such research. In order to make the subjects as similar to each other as possible (i.e, to control for a bunch of known and unknown variables), they picked only males within a particular range of age and IQ. They also picked right-handed subjects. This is important as there may be a corellation between lateral hand-use and lateralization of brain specializations. This is so important that they did not just ask the subjects – they gave them tests of handedness and quantified them before choosing the appropriate subjects.
I was always told to take this type of research with a grain of salt, and I am certainly in no position to judge the quality of their work from one set of computer-generated images they placed in the paper (everybody puts the best picture in there), so I will assume that peer reviewers were satisfied and that the data are sound.
One thing that worries me is that they did not compare “Stimulus” to “Non-stimulus”, but “Stimulus A” to “Stimulus B” to “Stimulus C” etc., a procedure that may produce statistically significant results from the same data that would not be significant if stumulus was compared to baseline (For instance male voice may stimulate one brain area and depress another, while the female voice would do exactly opposite, thus exaggerating the difference).
What can we learn from the data? Comparing responses to male and female voices (let’s ignore funny Latin names for the particular brain areas) shows that there is one area that is a little tiny bit more active when listening to a male voice and another area that is a tiny little bit more active when listening to the female voice.
There are some ideas what those two areas of the brain are doing, but all the extrapolations from the data are purely speculative. Female voices slightly increase the activity of the area of the brain that analyses sound. Male voices slightly increase the activity in the area that, among else, listens to one’s own voice, but is also generally involved in episodic memory (memory of personally experienced events including where and when they happened, not memory of things learned from books). Many “Just-So” stories can be weaved out of this.
The authors are very cautiously suggesting some possibilities along these lines, including the well-known fact that female voices are more complex (the iii category) thus may need more auditory processing, and the fact that male people have male voices thus are likely to compare male voices with their own (for whatever adaptive reason you can think of).
When gender-ambiguous and gender-unambiguous voices were compared, another brain area got a little bit more blood perfusion during the neutral-pitch stimulus. It may be that this area is involved in discrimination between natural and unnatural sounds, or just plainly paying attention. Gender of the speaker is an important aspect of perceived speech, so, if the gender is not easily recognized from pitch, the brain has to work a little bit harder to use the “other parameters” (iii) to figure out if the voice came from a male or a female.
The only place where authors allowed themselves a little bit of speculation was one sentence in the middle of this short paragraph at the end (note that “melodic quality” has nothing to do with singing):
Speech is an important carrier of identity. The features specific
to a speaker are ingrained in its acoustic parameters. We set out to
investigate the neural mechanisms that underpin the ability to
identify the gender of heard speech. Using the knowledge of
“gender-ambiguous” frequencies, we found that, in males, the
neuroperceptual factors which allow attribution of gender to heard
speech include those brain areas involved in interpreting intonation
(melodic quality) of speech for female voice identification and
those involved in mental imagery for perceiving a speaker as male.
Defining the brain basis for gender identification may help unravel
questions of between-gender differences in speech production that
allow organised complex social behaviour such as male-female
interaction for mate selection (Joseph, 2000; Semple et al., 2002).
From a clinical point of view, we propose that AVHs will be
associated with cortical activation in brain regions related to the
perception of the gender of the ”speaker” to whom the AVHs are
attributed. We predict that AVHs that comprise female voices will
involve brain areas used to perceive femininity in speech (i.e. right
STS). Likewise, we predict that the perception of AVHs assigned
as male in gender will be associated with brain areas used to
perceive male voices (i.e. precuneus).
Also, one of the criticisms on blogs was their choice of male subjects. When you do such tedious and expensive research and need as uniform sample as possible, you have to use one sex. They chose to do the males first (and immediately publish as soon as the data are analysed, preventing getting scooped) and do the females later (adding a second publication to their resumes):
Further work would primarily involve a repeat of the fMRI
paradigm on female subjects. Ultimately, we hope to implement the
paradigm on patients who experience AVHs, with the aim of using
our knowledge on the neural basis of gender identification from
heard speech to help disentangle and make sense of the multiple
brain regions associated with the actual experience of the
hallucinations (Woodruff, 2004).
You see? This is primarily about understanding schizophrenia and secondarily about understanding the brain, while sex differences and evolutionary considerations are to be chatted about later, over beer in a nearby Scottish pub.
In the same vein, look at this article from Psychology Today: The New Sex Scorecard
It starts OK:
It’s safe to talk about sex differences again. Of course, it’s the oldest story in the world. And the newest. But for a while it was also the most treacherous. Now it may be the most urgent. The next stage of progress against disorders as disabling as depression and heart disease rests on cracking the binary code of biology. Most common conditions are marked by pronounced gender differences in incidence or appearance.
Although sex differences in brain and body take their inspiration from the central agenda of reproduction, they don’t end there. “We’ve practiced medicine as though only a woman’s breasts, uterus and ovaries made her unique–and as though her heart, brain and every other part of her body were identical to those of a man,” says Marianne J. Legato, M.D., a cardiologist at Columbia University who spearheads the new push on gender differences. Legato notes that women live longer but break down more.
Do we need to explain that difference doesn’t imply superiority or inferiority? Although sex differences may provide ammunition for David Letterman or the Simpsons, they unfold in the most private recesses of our lives, surreptitiously molding our responses to everything from stress to space to speech. Yet there are some ways the sexes are becoming more alike–they are now both engaging in the same kind of infidelity, one that is equally threatening to their marriages.
Everyone gains from the new imperative to explore sex differences. When we know why depression favors women two to one, or why the symptoms of heart disease literally hit women in the gut, it will change our understanding of how our bodies and our minds work.
….then continues presenting both broadly accepted, and hypothetical, and very tentatively hypothetical, and very speculative and quite wrong ideas ALL as if they were accepted truths. Bad reporting, all over again.