Date: Tue, 20 Nov 2001 17:18:53 -0800 (PST)
From: Francis F Steen <steen@commstds.ucla.edu>
To: Martie Haselton <haselton@ucla.edu>
Subject: Re: Mona

Hi Martie,

Your letter may sufficiently clarify a central issue: do we predict male subjects are more likely to see the Mona Lisa as flirting? If the bias "operates regardless of whether a man is inferring the sexual interest of a woman in himself or in another third party man," this suggests we should get an effect.

Note, though, that this presupposes that the viewer construct a male spectator of the Mona Lisa -- perhaps the painter, perhaps the client (the person who commissioned and paid for the painting, perhaps the male gaze that in a broader sense is the target audience of the painting -- that is to say, a client may commission a painting of "his" woman in order to show her off to other men, and the painter may execute the painting with this purpose in mind. All of these male spectators are contemporaries of Mona Lisa (the woman and the depiction of her), and none of them are actually present in the painting. So I'm just pointing out that even this rather minimal inference of a male gaze involves a fair amount of imaginative construction of a fictive space.

Now, it may be that some version of this imaginative construction is in fact what any viewer, or any male viewer, tends to perform. If so, that would be interesting to document. It may be that the construction is largely unconscious -- or that the sheer existence of the representation of a woman immediately forces the inference that a male produced it. I actually think not -- surely there are representations of women that don't force the inference of a male gaze. But there may be a certain look that suggests a male gaze -- sort of, "she had the look of a woman that is being looked at by a man" -- and I suspect most people would say that the Mona Lisa clearly falls into that category.

This suggests a couple of things we could test:

1. We can show subjects a series of representations of men and women and ask them to rate the likelihood that the person represented was painted or photographed

• by a man
• by a woman

• for the enjoyment of the painter himself
• for a particular man, say the client, for private viewing
• for an audience of male social peers known to the client & painter
• for a general audience, at the time and now

We could approach this more explicitly in terms of probing imaginative immersion, asking subjects to rate the likelihood that:

• someone is looking at the Mona Lisa
• the person looking is a man
• the person looking is a woman

I like the second tack better; subjects may just feel they have no way of knowing the answers to the first set of questions, which really concern factual (EM) issues, while the second concern OM issues.

In short: if subjects indeed typically make the assumption that the Mona Lisa is being looked at by a man who is with her at the time and to whom she is subtly responding, then according to you we should expect sexual overperception in males. So we have a clear prediction!

The point about belaboring the exact conditions under which this happens is that I want to explicitly document the mental processes that go on during a viewing and not rely on our implicit understanding. The presence of the hidden male makes viewing the Mona Lisa more similar to watching a movie or reading a novel, in that these latter forms of entertainment precisely provide a fictive character with whom the viewer/reader can identify. So this new perspective on the Mona Lisa amounts to saying that we think male viewers of the painting identify with a man who is "in" the painting with the person depicted, looking at her, and that she sees him, but we don't. So are we agreed that that's what we're saying is going on?

A brief comment on the ambiguity issue. Is art valued for its ambiguity, or also for its realism? First, there's not necessarily a contradiction or contrast between ambiguity and realism, for two very different reasons.

First, a visual stimulus that is ambiguous could be very realistic. Take the analysis by Margaret Livingstone, the psychologist whose examination of the Mona Lisa first made me think of it in relation to your work (http://www.sscnet.ucla.edu/comm/steen/cogweb/CRP/MonaLisa/Livingstone_00.html). She argues that there is an actual visual ambiguity created by an artistic technique called "sfumato" or blurry. The effect of the sfumato technique in this case, she argues, is that when you focus on the mouth, you see it's not smiling, but when you look higher up, your peripheral vision interprets the shadows differently and makes it seem to the human visual system that she is smiling. Is this realism? That would seem to depend on what you're trying to depict: if you want to depict a face that appears to move -- and of course real faces move all the time, so that's surely a mark of realism -- then you could use a trick like that to produce the desired effect.

As a corrollary, blurring can in general be a realistic technique. A great deal of artistic technique has gone into making it appear that the subjects are moving; this marks the transition from the classic Renaissance style to Mannerism. Blurring outlines is the obvious way to achieve a sense of movement, since this mimics and taps into the eye's limited capacity for resolving a moving object. Advertizing today uses the same technique, as do cartoons. Is this unrealistic? Only if you assume that real objects always have clear boundaries and any blurring violates reality. If realism takes the way people actually perceive objects into account, then blurring is perfectly realistic. Under most circumstances, blurring won't cause any ambiguity. Livingstone may or may not be right that the perceived ambiguity of the Mona Lisa is due to the sfumato technique.

Second, ambiguity may be a result of higher-order inferences in the executive mode. This is really what made me think of your work. Strictly speaking, your work doesn't predict that people would find the smile ambiguous; it predicts that they will divide into two camps: those who thinks she's smiling (flirtatiously) and those who don't. If these two camps were to talk to each other, they might conclude that the smile is ambiguous -- and the explanation would be that men and women employ differing high-order inferences concerning the painting. Under the influence of the opposite perspective, perhaps an individual would also be able to experience both possible interpretations, and thus experience the painting as ambiguous. Again, this would be a purely higher-order cognitive effect, having nothing specifically to do with artistic techniques to manipulate the visual system.

Third, ambiguity may be a result of higher-order inferences in the organizational mode -- that's where my work comes in of course. The idea here is that any object is really ambiguous -- that is to say, it possesses or consists of a huge number of affordances, which can be drawn on for the construction of appropriate pretend situations. The meaning of an object isn't a fixed thing; it depends in any case on context, and in the case of the OM, it is wildly dependent on what the agent wants it to be. So the question here becomes more focused: an artist may be able to manipulate an object so that it becomes ambiguous at a specific level of abstraction, in a specific way. That would allow or entice viewers to explore a particular set of possibility spaces in a way that was very robustly supported by the artistic object. In this case, the ambiguity isn't the issue -- there is always an excess of ambiguity -- but the control fo ambiguity to be focused on a specific level that is interesting. So my suggestion is that the Mona Lisa is perceived to be ambiguous because the artist has created a facial expression that is in a sense uncommitted -- it could mean several different things. It isn't necessarily that Livingstone is right (though she could be); it would be enough if the smile is just plain old ambiguous, in a crisp and well-defined way, in the way that some real smiles are ambiguous. So in my perspective, the artist may have painted the Mona Lisa exactly true to life, completely realistically, and in fact so realistically that he captured that very ambiguous smile of hers. So in this case a picture could be extremely realistic and still highly ambiguous. It's just a matter of the level at which we locate the ambiguity.

So much for ambiguity and realism. Now about testing the predictions of these three theories? Let's start by considering them separately, and finish (finally!) by considering possible interaction effects.

First, consider scenario 1 (Livingstone). According to Livingstone's explanation, both male and femal subjects should consistently perceive the Mona Lisa to be smiling when they use their peripheral vision, and not to be smiling when they use their focal vision. There is a very simple way to test this hypothesis: eye-direction tracking equipment (Elizabeth Sowell says the Brain Imaging Center is getting it soon). Now, we may be able to improvise this test by the following experiment: Ask the subject to look at the mouth and report if she's smiling or not. Then ask the subject to look at her eys and respond to the same question. If it's purely an effect of the visual system, this should give us reasonably good results. We could even film the subject during the procedure. There should be no primary ambiguity; the ambiguity should all be an effect of trying to reconcile two non-matching interpretations. Each interpretation, however, should be reliably made by every individual.

There is no motivational component in Livingstone's work, so we might as justifiably predict that subjects will be annoyed by the painter's trick as that they enjoy it. Or they might be puzzled by it and feel something is wrong with their eyes. Her theory provides no explanation of the fame of the painting.

Second, scenario 2 (Haselton). According to your theory, we should also get two interpretations, and as under scenario 1, there should be no primary ambiguity. In this case, however, we expect the population to divide in two: those who think she's smiling, and those who don't. Of course, the effect isn't total, so even the initial division into two populations isn't clean. Would you expect some people to vacillate? As in scenario 1, there may be an increase in perceived ambiguity following reflection, in this case primarily reflection following exposure to opposite-sex participants in the study -- we could engineer that. Males should tend to think she's smiling, females not; after speaking together, they may both agree that it's ambiguous. It's unclear if this account explains anything about why the painting is famous -- flirting is of course enjoyable, but one might as easily imagine that men and women would become bitter enemies quarreling over whether she is smiling or not. According to your theory, the inference system itself is not subject to conscious awareness and the differing interpretations are not subject to conscious control, so it would be tempting in such a situation to conclude that men are from Mars and women are from Venus.

Third, scenario 3. According to my theory, all viewers, men and women, should perceive the Mona Lisa's smile to be ambiguous. They should not be confused by this ambiguity, and men should not disagree with women. The effect should be stronger in the OM than in the EM, if we do manipulations on those variables. That is to say, the Mona Lisa's smile should not simply be ambiguous; it should be a gate through which -- in the OM -- a flurry of different imaginative scenarios enter. There is a strong motivational component in the OM: her ambiguous smile should be experienced as intrinsically enjoyable.

Now, interaction effects. It's in fact quite possible that all three theories are right. This is perhaps a problem -- they're not really mutually exclusive. To some extent, of course, they are competing theories, in that the simple version yield very different predictions, as above. However, it's perfectly possible that ambiguity is enhanced by the artistic technique of sfumato, as Livingstone suggests; on the other hand, if I'm right about the OMH, then the eye-direction test should not produce clean results. People should think the smile was ambiguous, not just that it sometimes looks like she's smiling and sometimes not.

A case that is somewhat similar to the Livingstone analysis is the famous face/vase image: this is a certain type of ambiguity. If you focus on the black background, you see two faces facing each other. If you focus on the white space in between the faces, you see a vase. Nobody thinks the picture is maybe a vase and maybe two faces; it's clearly both. So according to this paradigm, in Livingstone's explanation, you should really have people being very confident at one point that she is not smiling, and at another that she is, and feel no need to put these together; it's just a trick painting. We would need some additional reason to think that people won't fully believe either perspective completely and attempt to fuse the two into one. Another objection to Livingstone's explanation is that the visual system may -- below the threshold of awareness -- detect an ambiguity in the peripheral visual field, but surely if the ambiguity completely disappears in the focal field, then that would dominate and the presentation that reaches consciousness would be unambiguously dour and non-smiling? This is a technical objection.

Livingstone could be right, however, in two ways (oh no!).

First, she could be right in that the sfumato technique gives rise to two differing interpretations, depending on whether you look at the mouth or not, and that these differing interpretations -- reaching consciousness -- are seized upon as interesting affordances for creating a possibility space exploring mental-state attribution in the OM. If she is right in this sense -- this is the combination of scenario 1 and 3 -- then scenario 3 predicts that people should *not* find the smile ambiguous in the EM. That is to say, if we frame the experiment to an EM situation, subjects should not see the smile as ambiguous; they should resolve it one way or the other, presumably to non-smiling, if this is what the focal vision tells you.

Second, Livingstone could be right in that the sfumato technique, having the perceptual effect she describes, perhaps provides information that would be selectively relevant for and utilized by men and women. Male subjects, given their proclivity towards sexual overperception, may make use of conscious or unconscious cues from the visual system that Mona Lisa is perhaps smiling to infer -- given their built-in bias -- that, darn it, she's a flirt. <g> Female subjects, provided with the same information, rejects the peripheral vision's contradictory suggestion of a smile, going with what good old reliable focal vision tells them is really going on: no smile. This is the interaction of scenario 1 and 2.

Now, what about interaction effects of scenarios 2 and 3? The obvious one is that, as we've spelled out before, men and women have different developmental trajectories, and an implicit curriculum would be predicted to favor appropriate and differing pretend practices. Specifically, the OMH would predict that males would be intrinsically motivated to explore possibility spaces filled with mating opportunities and to practice tasks -- such as discerning the level of a woman's interest -- that relate to such opportunities. In this practice, men should access evolved cognitive systems that are active in the EM, including sexual overperception. Thus, the interaction effect of scenarios 2 and 3 leads to the prediction that males will be more likely to infer that the Mona Lisa is flirting.

However, the meaning of the smile is different from the smile itself. Thus, the OMH does not predict that female subjects would be any less likely than male subjects to feel that the Mona Lisa may be smiling -- only that, in the interaction of scenario 2 and 3, that males would be more likely than females to infer that the woman depicted is sexually interested in the hidden male looking at her. A female subject may use the ambiguous smile to infer that the woman depicted is attempting to create the impression that she is sexually interested, or that she is really thinking about something entirely different, such as her lesbian girlfriend (the Gay and Lesbian group on campus sells a t-shirt of the Mona Lisa thinking, "I wonder if they know I'm lesbian?"). So the interaction of scenarios 2 and 3 suggests that men and women are equally likely to find it interesting to explore the possibility space opened up by the Mona Lisa's ambiguous smile,but they will explore different areas of this possibility space. Men will be drawn to explore the possibility that she is sexually interested in the hidden gazer to a greater degree than women will.

I like this final take on the interaction of our two theories.

Best,
Francis