Jan Lauwereyns, The Anatomy of Bias. How Neural Circuits Weigh the Options, MIT Press, 2010.

More and more often practicing scientists from the most diverse fields are writing books for general audiences with a view not only to communicate their own results or the state of the art in their field, but also to draw the more general implications of such findings for, say, our worldview. Whereas the former can be accomplished reasonably well by any competent scientist with a taste for writing, the latter will be more or less engaging depending on what the author has read beyond her discipline. Jan Lauwereyns is a cognitive neuroscientist and a remarkable poet who also enjoys reading across disciplines. And just as in 1621 The Anatomy of Melancholy provided an interdisciplinary survey on its topic, Lauwereyns presents his own anatomy as an "integrative account of the structure and function of bias as a core brain mechanism that attaches different weights to various information sources, prioritizing some cognitive representations at the expense of others" (p. xiv). There is much to praise in Lauwereyns' account, but I wonder to what extent it is really integrative. Let me explain why.

The original core of this book is mostly in the first two chapters, where the author presents the main findings of his own research, which hinge on his version of LATER, the standard model for the analysis of response time distributions generated in visual processing experiments. These experiments measure, on the one hand, the time it takes to a subject (usually monkeys) to convert a sensory stimulus into an eye movement according to the task assigned. On the other hand, they record the level of activity in a neuron (or group of neurons) that code for the relevant stimulus. In LATER the eye movement is understood as a result of a decision, modeled as a function of neural activity: when the function reaches a certain threshold, the eye moves. The function is defined by (1) the starting point of this activity, (2) its slope and (3) the variance of the activity.

In this framework, Lauwereyns claims that bias operates through changes in the first parameter that can be detected in two neural markers. First, the level of activity is significantly stronger in biased neurons than in unbiased neurons at all levels: i.e. for the coded stimulus and for noise alike. But even before the coded stimulus appears, and this is the second marker, the level of activity in biased neurons is also stronger, in a form of anticipatory processing. These superior levels of neural activity make the biased neurons reach the threshold at which decisions are made earlier than biased neurons.

Neural bias should be distinguished from neural sensitivity, the capacity to detect the right signal to act, which is measured by the second and third parameters in the model. The first marker for sensitivity is also a stronger level of activity, but just for the stimulus coded in the neuron, not for noise. The second marker is an enlarged ratio of response to the coded stimulus once it appears, through which the neuron can capture a broader range of signals.

Lauwereyns developed his interpretation of the LATER model namely in order to account for the activity of certain neurons observed in the caudate nucleus of monkeys in an experiment in which they had to make an eye movement to a position where a visual target had been briefly flashed shortly before. Certain caudate neurons, coding for positions where a reward had been obtained in previous blocks of the experiment, showed anticipatory activity before the visual target appeared (what Lauwereyns aptly calls wishful seeing). The monkeys, of course, could not predict where the next flash would come from. Further refinements of this experiment provided evidence that these neurons exhibited a reward-oriented bias of the sort described in the LATER model.

However, in the remaining five chapters, we do not find many more straightforward applications of LATER, but rather an informal examination of other experimental evidence in the light of this model. Hence, in chapter 3 Lauwereyns suggests that there are analogue biases and sensitivity mechanisms for fear (i.e., negative rewards) in the brain. In chapter 4, the author discusses the conceptual compatibility of his conception of bias with two widely studied heuristics in cognitive psychology, whose brain foundations remain as of today unexplored. In the remaining two chapters, Lauwereyns speculates on the more general brain architecture that could support LATER-like information processing. Chapter 5 is perhaps the more daring: Lauwereyns calls for the application of the theory of self-organizing processes to object representations in the brain. In chapter 6, he offers a conjectural model of competition of different neural networks in the brain that could account for evidence gathered in Stroop-like tests for monkeys. Chapter 7 contains the author's musings on the inevitability of bias in our species and how to tame it. The book closes with a Coda on the motivation for the book.

This is, of course, a quite partial summary aimed at capturing what, in my view, constitutes the book's thread or, at least, what I found most original or informative. My major complain about this anatomy is, precisely, how unbalanced it is in every other respect. The author is relatively systematic in the presentation of experimental evidence from his own field, but quite unmethodical in the discussion of everything else. As we have just seen, this is a book in which the author tries to generalize from experiments about certain brain mechanisms of visual processing in monkeys to bias in humans. This generalization requires a number of assumptions that the author clearly acknowledges: to name just one, about our cognitive architecture, how to model it and how to infer it from experimental evidence obtained in different species. Reading the book one gets to know Lauwereyns' views on these particular issues, but there is no introduction to any of them, much less a discussion of the alternatives. The author does not explore in depth positive research on biases in other disciplines, namely cognitive psychology, but rather handpicks examples without presenting their theoretical framework. The final discussion of the social consequences of our biases could have been improved with an examination of, e.g. different policies for fighting conflicts of interests in various domains (do scientific communities really fight biases the way the author thinks, for instance?) I do not think thus that this counts as an integrative anatomy of bias.

The Anatomy of Bias is intended instead as a more personal essay and we get to know more than our usual share about the author's life and tastes. I often got the impression that Lauwereyns digresses just because there is something he aesthetically likes, independently of whether he is right or wrong in the analysis (e.g. his occasional exegeses of Deleuze or Heidegger). Even if I am not particularly happy with this Anatomy, this is certainly a genre worth exploring, and Lauwereyns' attempt deserves all praise for trying to expand the scientific conversation beyond its usual borders.

{July 2010}
{Metapsychology online reviews}