Contribution
An encyclopedia entry on the family of theories that answer one question: what makes the difference between a conscious and an unconscious mental state of the same kind? The answer they share is a higher-order representation of the state. The entry sets out four versions and the objections each faces. It then turns to the claim that prefrontal cortex realizes the higher-order states. It argues for little. Its one general moral is that “different versions of a higher-order theory of phenomenal consciousness need to be kept distinct from one another” (§7). Its verdicts are that actualist HOT theory has not been refuted (§4) and that the issue of sufficiency is where “the battle for phenomenal consciousness may ultimately be won or lost” (§7.4).
Key insight
Conscious states are states we are aware of. Most kinds of mental state, perception and emotion included, also occur unconsciously, as in blindsight, absent-minded driving and the agnosic patient D.F. So whatever makes a state phenomenally conscious is something added to it, and higher-order theories say it is the subject’s non-inferential awareness of the state. The entry’s Transitivity Principle: “A conscious mental state is a state whose subject is, in some way, aware of being in it” (§2).
Assumptions
- Kinds of consciousness (§1). Creature consciousness against state consciousness; intransitive (awake) against transitive (conscious of). The entry advises dropping talk of transitive creature consciousness. Phenomenal consciousness (“feel”, after Nagel 1974) is set against functionally definable access consciousness (Block 1995). “Most theorists” hold that some states are access-conscious and not phenomenally conscious. The converse is disputed.
- A cognitive level of explanation (§2). Phenomenal consciousness “consists in a certain kind of intentional or representational content … figuring in a certain distinctive position in the causal architecture of the mind”. Mentality does not presuppose phenomenal consciousness, against Searle. Reduction is first to mentalistic terms, with a “second-step” reduction to neurophysiology often favoured.
- Unconscious perception exists (§2). Higher-order theorists “are united” on it. The evidence is the absent-minded driver (Armstrong); blindsight, with grasping in the blind field at “something like 80% or more of normal accuracy”; Milner and Goodale’s two visual systems; and D.F., at chance on the orientation of a slot she can post a letter through almost normally.
- Non-inferential awareness (§2). A desire learned of from a psychiatrist is a target of higher-order representation without being phenomenally conscious, so the awareness must normally be non-inferential.
Key results
- The first-order alternative (§2). For Dretske and Tye, phenomenal consciousness is analog, fine-grained content available to the first-order processes that guide thought and action. This explains ineffability: such contents “slip through the mesh of any conceptual net”. First-order theorists have three answers to the unconscious cases. They can deny the data (Dretske), draw the line by availability to belief rather than to movement (Kirk), or “bite the bullet” and hold blindsight percepts phenomenally conscious though not access-conscious (Block, Tye, Nelkin).
- The general thesis (§2). “A phenomenally conscious mental state is a mental state that is (or is disposed to be) the object of a higher-order representation of a certain sort”. The versions divide on whether the higher-order state is perception-like or thought-like, and on whether it must be actual or only disposed.
- Higher-order perception, or inner sense (§3). Armstrong, Lycan; after
Locke. Advantage: it gives purely recognitional concepts of experience,
which explain the explanatory gap and the conceivability of zombies by
their “conceptual isolation”. The entry gives three objections:
- There is no phenomenology of inner sense. The entry calls this question-begging, since inner-sense theory denies that first-order states have a phenomenology of their own.
- Inner sense could misfire, giving “It’s red” and “It seems orange” together, which never appears to happen.
- Carruthers’s complexity objection. The scanner would be “almost as sophisticated and complex as the visual system itself”, with no evolutionary story. “Lycan no longer holds HOP theory (Sauret & Lycan 2014)”: attention to first-order states now seems to him sufficient.
- Actualist higher-order thought (§4). Rosenthal: a state is conscious
when it actually causes, non-inferentially, a thought (generally itself
unconscious) that one is in it. Several points follow.
- Inference. Rosenthal rules out only conscious inference, so “unconscious self-interpretation is acceptable as a source of the conscious status of the states thereby attributed”. Noticing the spring in one’s step and inferring, unconsciously, that one is cheerful makes the cheerfulness conscious.
- Targetless thoughts. Rosenthal takes a suitably caused thought to suffice even with no target state. The entry calls both features optional for the theory.
- Regress. It is avoided because the thought is unconscious unless a third-order thought targets it in introspection (Figure 1).
- Dretske’s spot. Seeing the extra spot need not mean seeing that it is the difference.
- Richness. A rich scene would need many thoughts. The best reply is Dennett’s fragmentism, the “mere illusion of riches”, and the entry doubts it. Still, “it cannot yet be concluded that actualist forms of HOT theory have been refuted”, given change and inattentional blindness.
- “Light up”. How does thinking about a percept give it feel? Rosenthal points to wine-tasting and music courses, where new concepts change experience. The reply: oaky and oboe are first-order concepts, so the effect “appears to be a first-order phenomenon, not a higher-order one”.
- Dispositionalist higher-order thought, or dual content (§5).
Carruthers: a state is phenomenally conscious when it is available to
cause higher-order thoughts about itself, perhaps held in a short-term
store and made available by global broadcast.
- Consumer semantics. Content depends on the systems that consume a state, so availability to a “theory of mind” consumer makes each percept “at one and the same time” an analog representation of green and of seeming green.
- Prosthetic vision. The independent illustration is Bach-y-Rita’s tactile-visual substitution. With the camera under their control, subjects’ tactile patterns came to represent objects in space. “Here everything on the input side remains the same as it was when subjects first began to wear the device; but the planning and action-controlling systems have learned to interpret those states differently.”
- Misrepresentation. It cannot happen, since the higher-order content is “parasitic” on the first-order one.
- Costs. It requires rejecting pure input semantics. Rosenthal objects that dispositions do not make us aware of anything. The objection that it makes feel non-categorical counts “equally against any representationalist theory”.
- “Carruthers no longer holds dispositional HOT theory or, for that matter, any form of higher-order theory”, and defends first-order representationalism (Carruthers 2017).
- Self-representational and hybrid theories (§6). These deny that the
higher-order state is a distinct state: the relation is “constitutive,
or internal”.
- Identity versions: Carruthers’s dual content, and Van Gulick’s HOGS.
- Part–whole versions: Gennaro’s wide intrinsicality view, Kriegel, and Picciuto’s quotational QHOT.
- Kriegel posits a “ubiquitous conscious (but inattentive or peripheral) self-awareness” with every outward conscious state, which Gennaro doubts.
- Kriegel requires that the integration be one “resulting from a psychologically real process (as opposed to a theorist’s definition)”.
- A reconstruction on the model of the duck–rabbit: conceptual templates matched against analog input yield one state that represents the lines, a rabbit and “that one is experiencing something rabbit-like”, which so becomes “self-presenting”.
- Targetless representations (§7.1). Inner sense and actualist HOT
face it; dispositional and self-representational views do not.
- Lycan and Rosenthal accept the consequence. Lycan cites morphine. Rosenthal cites dental patients with dead nerves who feel pain until told.
- A critic’s alternatives. A vivid imagining of pain, or “a top-down expectation of pain causes a first-order experience of pain”, placebo’s opposite.
- Gennaro. The first-order state must exist.
- Brown’s HOROR makes the theory non-relational. Rosenthal (2022) objects that “[i]t’s the first-order state that’s qualitatively conscious”.
- Block (2011). A thought that one is in pain should not be expected to have pain’s “negative valence and high-arousal properties”, so it is not sufficient for felt pain.
- The rock (§7.2). Thinking about a rock does not make it conscious. Replies: a rock is not a mental state (Lycan), or cannot be part of a complex higher-order state (Van Gulick, Gennaro).
- Animals and infants (§7.3). Higher-order thought seems to need mental-state concepts that most animals may lack. Gennaro’s replies are simpler indexical higher-order thoughts, and comparative evidence of metacognition and mindreading: rhesus monkeys, scrub jays re-caching, episodic memory. Van Gulick’s implicit higher-order representation over-generates: “It is implicit in the behavior of any creature that drinks, for example, that it is thirsty”. The theory “does not say that having conscious states requires having introspective states”.
- Sufficiency (§7.4). Creatures with the right higher-order representations but no phenomenal consciousness seem conceivable. The reply, shared with first-order theorists, is that reductive explanation needs only “good reason to think that the explained properties are constituted by the explaining ones, in such a way that nothing else needed to be added to the world”. “But this is hotly contested territory.” Somatoparaphrenia and thought insertion, where states are felt as not “mine”, run against the HOT claim that the thought “I am in mental state M” accompanies “each and every conscious state”.
- Prefrontal cortex (§8).
- For. Prefrontal and parietal activity is higher in conscious than unconscious conditions, “often even when performance capacity is controlled for”. In Lau and Passingham (2006), subjective reports were impaired with discrimination intact and tracked dorsolateral prefrontal cortex. In Rounis et al. (2010), TMS to prefrontal cortex affected “meta-visual awareness” but not first-order performance. The prediction: disrupting prefrontal cortex “should affect or eliminate sensory experiences without affecting performance”.
- Against (localists). Little prefrontal activity while absorbed in a film (Goldberg et al. 2006). Experience is not eliminated by extensive prefrontal damage or lobotomy. Prefrontal activity may come from report demands.
- A middle position. Gennaro locates higher-order states outside prefrontal cortex: anterior cingulate, temporo-parietal junction, parietal cortex.
- The verdict. The debate “continues”, now with adversarial collaborations (Ball 2019). The entry adds that if prefrontal activity is required, “it is unlikely that most animals or even infants are conscious”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Most kinds of mental state, perception included, occur both consciously and unconsciously | moderate as reported | blindsight, D.F., the absent-minded driver, §2; first-order theorists contest whether these are unconscious |
| C2 | What makes a state phenomenally conscious is its being the object (or disposed object) of a suitable non-inferential higher-order representation | the thesis surveyed; argued, not settled | §2; the entry concludes no verdict |
| C3 | Actualist HOT theory has not been refuted by the richness objection | weak: a judgement that the matter is open | fragmentism, change and inattentional blindness, §4 |
| C4 | Concept learning that changes experience (wine, music, the duck–rabbit) is a first-order effect, not evidence for HOT | moderate as argument | §4, §6 |
| C5 | Dispositionalist and self-representational views avoid targetless representations; inner sense and actualist HOT do not | strong, by the structure of the views | §7.1 |
| C6 | An implicit higher-order representation over-generates consciousness to every creature | moderate as argument | the thirst example, §7.3 |
| C7 | Prefrontal activity accompanies conscious against unconscious conditions with performance matched | weak here: reported, and contested as report-driven | Lau & Passingham 2006; Rounis et al. 2010; §8 |
| C8 | Whether higher-order representation is sufficient for phenomenal consciousness turns on what reductive explanation requires | the entry’s framing | §7.4 |
Concepts
- Transitivity Principle (TP): a conscious state is one whose subject is “in some way, aware of being in it” (§2).
- Higher-order perception (HOP), or inner sense: a faculty that scans first-order perceptions and produces analog higher-order representations of them (§3).
- Actualist HOT: consciousness by actually causing a non-inferential higher-order thought (§4).
- Dispositionalist HOT, or dual content: consciousness by availability to higher-order thought, which under consumer semantics gives each percept a second, “seeming” content (§5).
- Self-representational theory: one state represents the world and itself, by identity or as part and whole (§6).
- Targetless (empty) HOT: a higher-order representation with no first-order target (§7.1).
- HOROR: higher-order representation of a representation, whether or not the target exists (Brown 2015; §7.1).
- Localist and globalist: whether some conscious states need only local sensory activity, or need widespread activation (§8).
Connections
The entry builds on Armstrong (1968), Rosenthal (1986, 2005), Lycan (1996), Carruthers (2000, 2005), Gennaro (1996, 2012), Kriegel (2009) and Van Gulick, and engages Dretske (1993, 1995), Tye (1995), Block (1995, 2007, 2011), Levine and Chalmers (1996). Of these the record holds Block 1995 (LIT-643), Block 2007 (LIT-646, which §8 cites for the definition of a neural correlate), Block’s “Advertisement for a Semantics for Psychology” (1986, LIT-648, named among the consumer semantics) and Nagel 1974 (LIT-096). Block’s own comparison of higher-order, workspace and biological theories (LIT-632) is not cited. Its companion entry is Lycan’s on representational theories (LIT-1355), read alongside. Lycan there sends awareness to this entry, and this entry reports that Lycan has since given up inner sense.
Three things in the text should be noted.
- The adversarial collaborations. §8 mentions them only through Ball (2019) and reports no result. The COGITATE test of global workspace against IIT (LIT-973, 2025) is not cited, though the entry was revised in January 2026.
- The figure label. Figure 1’s long description is labelled “figure 6”.
- Two pointers that do not check out. §4 says “As noted earlier” of Rosenthal’s reading of the non-inferential condition, which is not noted earlier. §7 sends the reader to “section 3” for the framing of objections in terms of HOT theory.
Bearing on the record
- CLAIM-411: the claim’s premise, and the most developed answer to
it. The entry’s motivation (§2) is the claim’s premise in the same
cases: blindsight, automatic motor control (D.F.’s sensorimotor
system) and the absent-minded driver.
- Both defeater branches are occupied. The higher-order thesis is a stated criterion of the kind the claim’s first branch asks for. The first-order theorists who “bite the bullet” (Block, Tye, Nelkin) hold the second branch, that blindsight percepts are experiential.
- Neither branch is won. The higher-order criterion faces the same
two failures the claim predicts for any candidate:
- Necessity. Animals and infants (§7.3), and the localist lesion evidence (§8).
- Sufficiency. The conceivable creature with the right higher-order states and no experience (§7.4), and the rock (§7.2).
- The evidence is report-tied. The empirical case for it (§8) contrasts “conscious versus unconscious conditions” by report, and localists reply that the prefrontal activity is report. So it does not yet meet the claim’s requirement that the criterion be “independent of report”.
- Over-generation. Van Gulick’s implicit version (§7.3) does what the claim’s defeater forbids: it lets the verdict spread to “every creature (no matter how simple)”.
- The claim stands, and this sharpens it. The leading named candidate has been abandoned by two of its main sources (Carruthers, Lycan) and is maintained by a third.
- CLAIM-412: three near-relatives, each with a difference.
- Interpretation that confers consciousness. On Rosenthal’s actualist theory, “unconscious self-interpretation is acceptable as a source of the conscious status of the states thereby attributed” (§4). That is the literature’s closest case of interpretation making a state conscious. But the interpretation is a self-ascribing thought, a higher-order judgement, which is the sense the claim puts aside. The claim treats judgement as “a higher-order operation within an interpretive system” (A1 of the exchange, not this entry). So the claim cannot adopt it without reversing its part 2.
- Content constituted by how a state is consumed. Carruthers’s dual content (§5) is constitutive, not causal: availability to a theory-of-mind consumer makes the percept “at one and the same time” a representation of seeming. His prosthetic-vision illustration holds the input fixed while consumer systems “have learned to interpret those states differently”, and the content changes. That is the claim’s picture for content: significance fixed by the organization that uses a signal. Two differences matter. The experiential dimension is conferred by a mentalizing system, which brings in the animals objection. And Carruthers has abandoned the view.
- Realized, not described. Kriegel’s requirement that the integration result “from a psychologically real process (as opposed to a theorist’s definition)” (§6) is the claim’s answer to Negro: realized, not merely described, organization.
- The standard of explanation. The entry’s reply on sufficiency (§7.4) is the claim’s standard for an identity, that “nothing else needed to be added to the world”. The entry says that this ground is where the issue “may ultimately be won or lost”. It does not settle it.
- QUESTION-254: the question’s second contrast is contested. The question separates constituting a perspective from representing one. Higher-order theories make the experiential difference consist in representing the state: the higher-order content gives “a dimension of seeming or subjectivity” (§3, §5, §6). On these views a representation of one’s own state (nearer TERM-072’s sense 2) is what makes a state experiential. A self-representational theory goes further: the state is “self-presenting”. So the question’s separation is a substantive position against this family, and an answer to the question owes a reply to it. The question’s overgeneration test applies to the family as it applies to the claim (§7.3).
- TERM-072.
- The bridge from sense 2 to sense 3. Higher-order theory is a standing attempt to make the bridge the TERM says must not be built into a definition: represented self-awareness (sense 2) as what makes phenomenal consciousness (sense 3).
- Berger’s view. Berger (2014) makes consciousness a property of persons, “how my mental states appear to me”, not of states (§7.1). That is nearer a perspective-level reading.
- TERM-073. The entry uses “interpret” in two senses, neither
the TERM’s.
- Rosenthal’s “self-interpretation” is inference from one’s own behaviour to one’s own state.
- Carruthers’s consumer systems “learned to interpret those states differently”. That is a change in how a state is used, and it is the nearer one.
- CLAIM-355. The pathologies of §7.4 bear on it. The HOT content “I am in mental state M” builds an owner into every conscious state. In somatoparaphrenia and thought insertion the ownership attribution fails while the experience stays. That fits CLAIM-355’s location of the owner in a representation that can fail without the binding failing. It is a problem for HOT theory, and the entry cites Gennaro’s discussion of it, which was not read.
- QUESTION-048, THEORY-094, THEORY-098. §8 restates the prefrontal dispute that THEORY-098 takes a side on. It lists the same localist evidence (lesions, film-watching, report-driven activity) and the higher-order evidence the THEORY does not discuss (Lau and Passingham; Rounis et al.). It also reports the report confound THEORY-094 states: “it is crucial to show that the activity detected in the PFC during imaging studies pertains primarily to attention and report, not consciousness per se”. It adds nothing that settles either, and it does not report COGITATE.
- THEORY-577 and THEORY-633. Inner-sense theory claims to explain the explanatory gap and zombie conceivability through purely recognitional concepts (§3). That is an attempt at the meta-problem THEORY-577 poses. Rosenthal’s “Consciousness is about mental appearance” (§7.1), with targetless thoughts sufficing, is close to THEORY-633’s quasi-phenomenal reading but keeps phenomenal consciousness real.
- THEORY-766 (expectation). The dental case (§7.1) has the shape of that THEORY’s open question. A reported experience may be produced by the stimulation, by a thought about it, or by “a top-down expectation of pain” that causes a first-order experience. The entry does not decide among these.
- THEORY-879, filed from the companion entry, says that a sensory quality needs awareness to be experienced. This entry is the account of that awareness it defers to. The entry also grants, for the driver, that one “may be possible to get oneself to believe” the perceptions are phenomenally conscious, so the case that THEORY rests on is not settled here either.
- No THEORY filed. The entry argues for no thesis of its own beyond the moral that versions must be kept apart. A THEORY on higher-order thought should come from a primary source (Rosenthal 2005; Lau and Rosenthal 2011; Brown, Lau and LeDoux 2019), none of which the record holds.
- No instruction for machine-learning practice; nothing for the anthology.
Limitations
- A survey maintained since 2020 by Gennaro, whose own views (wide intrinsicality, animal metacognition, the “centrist” neural location) get the last word in several sections.
- Every empirical result is reported, with no effect sizes, and none was read here.
- §8 reports the prefrontal debate up to about 2023 and does not report the published results of the adversarial collaborations it mentions.
- Emotion gets one paragraph (LeDoux and Brown’s HOROR for fear, §8), and affect enters only through Block’s point about pain’s valence (§7.1).
Open questions
- Is there a criterion for “higher-order” that does not over-generate (Van Gulick’s thirst) or under-generate (animals without mental-state concepts)? §7.3 leaves it open.
- What would separate Rosenthal’s targetless dental pain from a top-down expectation that produces a first-order pain? The entry names both and no test.
- Does disrupting prefrontal cortex remove experience, or only metacognitive report of it? This is the theory’s own prediction (§8). Rounis et al. report the second, and whether the first follows is open.