In his 1928 book Der Logische Aufbau der Welt (sometimes translated as The Logical Construction of the World), Rudolf Carnap attempted to show that all meaningful scientific terms could be defined in phenomenalistic language (language describing conscious experiences), in which case knowledge of phenomenal truths would suffice for knowledge of all scientific truths.Footnote ¹ Carnap's Aufbau was a landmark of phenomenalism and of logical empiricism. It was also a failure, in the judgment of many philosophers. Although much admired for its rigor and ingenuity, the Aufbau is generally taken to fall short of achieving its overarching ambitions, including especially that of analyzing physical truths in purely phenomenalistic terms.Footnote ²

However, something close to Carnap's project is pursued in David Chalmers' new book, Constructing the World, the title of which is meant as an homage to Carnap's work. What Chalmers takes from the Aufbau is not Carnap's phenomenalism or empiricism, but rather the idea that there is an important subclass of truths, knowledge of which would suffice for knowledge of all truths.

Constructing the World is a work of major philosophical importance that will be of interest to philosophers of just about every stripe. It is extremely ambitious. Here is a partial list of topics where substantive philosophical theses are advanced (sometimes tentatively): the analytic/synthetic distinction, the a priori, Fregean theories of meaning, content externalism, the role of conceptual analysis within philosophy, Bayesian epistemology, perception, color, normativity, fundamental ontology, quiddities, metaphysical grounding, negative truths, the laws of nature, scientific reduction and the unity of science, structural realism, space and our spatial concepts. And yet even with all the territory it ranges over, the argumentation is consistently careful and rigorous.

The book is divided into eight chapters along with seventeen (!) excurses that are interspersed throughout. In the first two chapters Chalmers introduces the notion of scrutability and various scrutability theses whose elaboration and defense make up the core of the work. This includes Conditional Scrutability, the thesis that there is a compact class of truths such that for any truth p whatsoever, a Laplacean intellect – that is, a hypothetical being who comprehends all the truths in the compact class, and who possesses ideal reasoning capability to derive further truths from that class – would be in a position to know that if the truths in the class obtained, then p. Also, A Priori Scrutability, the thesis that there is a compact class of truths such that for any truth p, a Laplacean intellect would be in a position to know a priori that if the truth in the class obtained, then p.

It bears emphasis that the class of truths serving as the scrutability base must be compact, meaning that they use expressions from only some small number of families of expressions. For much of the book, Chalmers assumes as a working hypothesis that the scrutability base consists in what he calls PQTI truths: Physical truths (whose expressions are taken from the science of physics), Qualitative truths (whose expressions describe conscious experiences), Indexical truths (whose expressions include terms like ‘I’ and ‘now’), and a That's-all truth saying that there are no positive truths beyond those scrutable from the physical, qualitative, and indexical truths. The PQTI base is compact in that it uses expressions from only those four families. Or at least that's the idea – below, we will return to compactness and the that's-all truth.

Provisionally adopting the PQTI base, A Priori Scrutability amounts to the thesis that a Laplacean intellect that knew all the physical, qualitative, and indexical truths about our world, together with the further that's-all truth, would thereby be positioned to know absolutely all the truths of our world. (Mutatis mutandis for Conditional Scrutability.) For example, such a being would be positioned to know the truth that there is an armchair in the living room. For, although this is not itself a PQTI truth, it is something a Laplacean intellect could ‘scrute’ from the PQTI truths.

A good portion of Chapter 3 is spent defending the Conditional Scrutability thesis by appealing to a ‘Cosmoscope’, a hypothetical device that provides ordinary (non-Laplacean) subjects with exhaustive physical, qualitative, and indexical information about arbitrary regions of the world. Using a Cosmoscope as a kind of virtual reality device, you could undergo the sort of visual experiences one would have upon entering the living room – visual experiences as of an armchair. Now, this qualitative information taken by itself is compatible with skeptical scenarios on which there is in fact no armchair in the living room – there is just an elaborate armchair hologram, say. But in addition, using a Cosmoscope you could learn that there is physical matter in the living room that occupies just the spatial position that the armchair is supposed to occupy, causally interacting with the world in just the way armchairs are supposed to, and so on. Now there seems to be no room left for skeptical scenarios – the Cosmoscope has positioned you to know the truth that there is an armchair in the living room. We can thus conclude that this truth is conditionally scrutable from the PQTI base.

In Chapter 4, Chalmers draws in part on conclusions established by the Cosmoscope argument to defend the A Priori Scrutability thesis. The notion of ‘frontloading’ plays a key role. Roughly, the idea is that if a given proposition p can be known on the basis of empirical evidence e (that is, on the basis of knowing e by virtue of possessing empirical warrant for it), then the following conditional can be known a priori: if e, then p.

Chalmers' project in the book relies heavily on the notion of the a priori and on a form of conceptual analysis. In Chapter 5 he rebuts Quine's influential critiques of these notions, and in connection Quine's attack on the analytic/synthetic distinction.Footnote ³ Chalmers' argument here, which was earlier set out in his paper ‘Revisability and Conceptual Change in “Two Dogmas of Empiricism”’, makes use of a Bayesian framework.Footnote ⁴

In Chapter 6 Chalmers considers whether the scrutability base of PQTI needs to be expanded – that is, whether this base is rich enough to allow a Laplacean intellect to scrute truths about mathematics, ontology, intentionality, modality, and yet other ‘hard cases’. Chapter 7 is devoted to the opposite question of whether the PQTI scrutability base can be shrunk down. Both chapters are fairly exploratory in nature. Chalmers is more interested in setting out the possible views than in defending any particular view at great length. Chapter 8 takes up the question of whether the scrutability base can be made principled, meaning roughly that there is some non-arbitrary condition satisfied by all base expressions. These are the chapters in which the discussion becomes especially wide-ranging, touching on many of those philosophical topics that I listed near the beginning of the review.

With my remaining space I want to sketch (without necessarily endorsing) an objection to Chalmers' views. The sort of objection it will be most illuminating to consider is one that threatens the scrutability theses that are at the heart of the work. To focus discussion, I will limit my attention to A Priori Scrutability.

My objection concerns the that's-all truth. Chalmers justifies his inclusion of this truth within the scrutability base by appealing to negative truths like ‘There are no ghosts’. Negative truths appear to be inscrutable from PQI truths taken by themselves (or, for that matter, from positive truths more generally) since there are worlds exactly like ours with respect to PQI truths but which contain ghosts in addition. Given that there are such worlds, a Laplacean intellect that knew only the PQI truths of our world would not thereby be positioned to know whether our world did or did not contain ghosts.

Once the point is appreciated, it becomes plausible that there are also certain positive truths that are inscrutable from PQI truths – inscrutable, because a Laplacean intellect would need to know certain negative truths in order to scrute these positive truths. For example, Ned Block and Robert Stalnaker suggest in effect that the truth ‘Heat is molecular motion’, which is not intuitively negative, is inscrutable from physical truths alone.Footnote ⁵ On their view, in order for a Laplacean intellect to scrute this truth she would also need to know the negative truth that there is no ‘ghost heat’, a nonphysical substance that plays heat's functional role even better than molecular motion does, and so which is the better candidate to be identified with heat.Footnote ⁶

So then, how to render such truths scrutable? One option – an option Chalmers does not take – would be simply to include the entire class of negative truths, N, within the scrutability base. Chalmers could then reply to Block and Stalnaker that ‘Heat is molecular motion’ is a priori scrutable from the physical truths taken in conjunction with the negative truths, including the truth ‘There is no ghost heat’. More generally, Chalmers could then hold that all truths are a priori scrutable from PQNI truths.

The problem with this option is that it would render the scrutability base incompact. For any family of expressions there is, there will be some negative truth that uses expressions from that family: ‘There is no ghost heat’, and ‘There is no largest prime number’ and ‘I am not in pain’ are all included in N despite the diverse vocabularies they employ. In that case, even if PQNI truths are such that for any truth p, a Laplacean intellect would know a priori that if PQNI, then p, this would not be good enough to establish Chalmers' A Priori Scrutability thesis. It would not be good enough because PQNI truths do not comprise a compact class, as that thesis explicitly requires.

Again, Chalmers does not take the option we have been considering; he instead includes the single that's-all truth within his scrutability base.Footnote ⁷ But it is not immediately obvious this genuinely solves the problem just outlined. For, what does knowing the that's-all truth consist in? You can't just see or feel that that's all, in the way you can see or feel an armchair.

Here is one possible proposal: knowing the that's-all truth consists in knowing that various arbitrary nonexistents do not in fact exist – for instance, knowing that there are no ghosts, or unicorns, or etc. But in that case, knowing the that's-all truth consists in knowing various arbitrary negative truths, which use expressions from all varieties of families of expressions. If so, then even if Chalmers is right that all truths are a priori scrutable from PQTI truths, this would not be good enough to establish his A Priori Scrutability thesis. It would not be good enough because PQTI truths would not comprise a compact class of truths, as the thesis explicitly requires.

I will not try to determine here whether this general line of objection is decisive; various moves and countermoves would need to be considered. Instead, I will close by reemphasizing that Constructing the World is a fantastic book. Working through it has done much to clarify my own thinking, both on points where I am inclined to agree with Chalmers but also on points where I am inclined to disagree.