1 Introduction

It has been argued that metacognitive experiences are evidence for cognitive phenomenology (Dorsch, 2016; Goldman, 1993; Robinson, 2005; Smithies, 2013) or against cognitive phenomenology (Arango-Muñoz, 2019; Carruthers & Veillet, 2011; Prinz, 2012). Several authors have defended embodied affect views of tip-of-the-tongue (TOT) and other metacognitive experiences (Arango-Muñoz, 2019; Dokic, 2012, Dorsch, 2023, Proust, 2013, 2015) according to which such experiences are nonconceptual, lending credence to arguments that metacognitive experiences do not include cognitive phenomenology. In this paper, I argue for a contrasting cognitive view, according to which the phenomenal character of TOT is in part constituted by conceptual representations. The paper begins by clarifying key concepts and claims. Next, I examine recent philosophical arguments that metacognitive feelings are nonconceptual embodied affects. I show why none of these arguments give us reason to reject a view of TOT according to which the phenomenal character of such experiences is in part constituted by conceptual representations, in other words, none of these arguments show that TOT does not include cognitive phenomenology. I then review the history of the psychological study of TOT and demonstrate how empirical data from the science of language production supports a cognitive view, as well as discussing contrasting heuristic-metacognitive views. I end by assessing the relative strengths of the Cognitive view of TOT in comparison to alternatives. Throughout the paper I present several avenues of clarification that defenders of embodied affect views would have to address to situate their views within the cognitive phenomenology debate.

2 Cognitive Phenomenology

Is the conceptual content of my thinking, the meaning of words spoken by others, and the conceptual identification of objects I perceive a constitutive part of my conscious experience? Some philosophers deny this, arguing that the phenomenal character of our experience only contains perceptual states and their inner analogues, such that when thinking or understanding, you experience only perceptions, mental visual imagery, or sentences in inner speech. In contrast, proponents of cognitive phenomenology answer affirmatively, arguing that our experience is not exhausted by merely sensory features, instead, cognition makes a constitutive contribution to the phenomenal character of experience.

The sense of “consciousness” at issue is phenomenal consciousness. Phenomenally conscious states are those that we experience, there is “something it’s like” to be in such a state (Nagel, 1974). More specifically, cognitive phenomenology is focused on the phenomenal character of our conscious experiences, and to refer to the phenomenal character of an experience is to refer to those features that we experience when we are in such a state (Siewert, 1998).

Let’s consider the views of those who deny that there is cognitive phenomenal character. Lormand (1996) provides a list of states that he argues can be phenomenally conscious, which he labels the “Qualitative Quartet:”

  1. 1.

    Conscious Perceptual Representations

  2. 2.

    Conscious Bodily Sensations

  3. 3.

    Conscious Imaginings

  4. 4.

    Conscious Streams of Thought Composed of Inner Speech and/or Imagery

Michael Tye and Briggs Wright (2011) follow Lormand in restricting phenomenality to these states, with one addition:

  1. 5.

    Emotional Experiences

Whatever conscious experiences we might enjoy, according to Tye and Wright, they will always and only be some combination of the above quintet of states. To understand exactly how Tye and Wright deny cognitive phenomenology, it is important to note that the phenomenal character of these states “is not inherently conceptual or cognitive” (2011, p. 329), instead, they are merely sensory. Roughly all deniers of cognitive phenomenology share the position that the above quintet of merely sensory states are all that is ever found in phenomenal character.

Proponents of cognitive phenomenology argue that the above list of phenomena is not sufficient to explain the phenomenal character of our cognitive experience. For example, Michelle Montague argues:

It is not possible to give an adequate account of what a conscious thought consists in … without appealing to cognitive phenomenology … the notion of the ‘phenomenological’ or ‘phenomenal’ must extend beyond the sensory. Sensory phenomenological properties are not the only kinds of phenomenological properties there are. (2016, p. 175)

Proponents of cognitive phenomenology are committed to expanding phenomenality beyond the Quintet, but further clarifications are needed. Because cognitive phenomenology does not concern just any sort of phenomenological properties that extend beyond the sensory, but properties that are robustly cognitive. A popular way of narrowing the scope of concern is to understand the cognition at issue in terms of the exercise of conceptual capacities, i.e., cognition that includes the exercise of concepts, which are essentially involved in our capacities for classification, rational inference, and linguistic understanding.

Note that the conceptual activity relevant for cognitive phenomenology is conceptual activity that makes a constitutive contribution to the phenomenal character of our experience, not a merely causal contribution (Carruthers & Veillet, 2011, pp. 36–38). This is only to say that to show there is cognitive phenomenology it would not be sufficient that some unconscious conceptual activity causes a modification in our sensory phenomenology, instead, the conceptual activity itself or the conceptual representations involved in it must be a part of—must constitute, at least partially—the phenomenal character of conscious experienceFootnote 1. Consider that deniers of cognitive phenomenology grant that there are cognitive causal influences on sensory phenomenal character, for example, that my understanding a language will influence how I hear that language spoken by influencing how words are grouped into phonemic units (Tye & Wright, 2011, p. 337) or by altering the accompanying visual imagery I experience (Prinz, 2012, pp. 160–163). Likewise, proponents grant that there are sensory causal influences on cognitive phenomenal character, for example, that visually or auditorily perceiving words may bring about my experience of understanding their meaning (Siewert, 2011, pp. 251–258, Strawson, 1994, pp. 5–7, Dorsch, 2017). In both types of examples, the proponent or denier, respectively, agrees that there are sensory or cognitive causal underpinnings of such experiences, but because these causes are external to the phenomenal character of the experiences, they are not considered part of their phenomenology. What proponents and deniers centrally disagree about is what sort of representations can constitute phenomenal character, i.e., they disagree about whether the consciously experienced aspects of our mental states are merely sensory, or whether they are also made up of cognitive features.

In sum, proponents of cognitive phenomenology argue for the following claim:

Cognitive Phenomenology (CP): the phenomenal character of some conscious experiences is constituted (in part) by conceptual representations.

My aim in this paper is to examine the controversy over whether TOT is an example of such an experience. I will argue that a proper interpretation of TOT supports CP.

3 Cognition and Metacognition in TOT

A TOT occurs when one is trying to retrieve some name or word from memory, one fails to retrieve it, and crucially, the subject experiences knowing the word, they can think about the object or person this word refers to, and they know what they mean despite being unable to say it. The fact that this experience includes the contradictory elements of both knowing and not knowing is what makes it feel so acutely strange and frustrating, and how to explain what one experiences that gives rise to that seeming-to-know is where the controversy begins. Clearly, what one is not conscious of when experiencing TOT is a phonological representation of the word, but what one does experience is less obvious. The key issue for this paper will be examining those positive qualities that one does experience in a TOT state, in particular, whether the phenomenal character of TOT includes only non-conceptual embodied affects, or whether TOT phenomenology also includes conceptual representations, thereby supporting CP.

What is the difference between a “cognitive” and “metacognitive” mental state?

A basic way of conceiving this distinction in psychology is that a metacognitive state is one that is evaluative of some other cognitive state or process. For example, we might describe one’s memory retrieval processes as “cognitive” states, while we might call some mental state that is evaluative of such cognitive states as “meta-cognitive.” For examples of cognitive representations that one might experience in TOT, consider recognizing a person and forgetting her name, but recalling some of her semantic features (Dave’s wife, works at the libraryFootnote 2). Or being unable to retrieve the name of an object but knowing it’s semantic properties (metal tool, used to open a tight lid). Both cases are first-order cognitive states constituted by semantic representations. In the case of metacognitive representations, examples would include feeling like one is close to recalling a word, which is a second-order evaluation of one’s first-order memory, or expressing in inner speech a metacognitive judgement that you are in a TOT state in reaction to the feeling that you are in one, which is a second-order evaluation of a first-order state. An experience being a first-order or a second-order state of these types does not by itself matter for debates concerning CP, rather, it is how the phenomenal character of these states is constituted that matters for CP. It is simply a contingent feature of the current philosophical literature that the metacognitive views that have been defended understand those metacognitive states to be non-conceptual embodied affects, and so it is important for me to analyze whether these views concerning metacognition support the exclusion of conceptual representations from the phenomenal character of TOT experience.

4 Metacognitive Feelings as Embodied Affects

Proust (2015) understands metacognitive experience to involve metacognitive feelings, where these feelings are understood to be nonconceptual, embodied affects. Consider how Proust defines the term “feeling:”

‘Feeling’ denotes a reactive, subjective experience with a distinctive embodied phenomenal quality and a formal object, which may or may not coincide with the embodied experience. Feelings typically express affect and valence in sensation. (2015, p. 2)

Proust explains that the metacognitive feelings involved in TOT “are conveyed by the intensity of the activity in the tongue muscle, and by the affective quality of TOT” (2015, p. 16).

Proust’s claims regarding the embodied aspects of TOT experience are dubitable on introspective grounds, although Proust does not seem to base her argument on introspective evidence. Introspectively, it does not appear that the TOT experience is located in the body. I experience TOTs within my mind; they have the feel of something mental, rather than bodily. I have never experienced a TOT as localized in my tongue muscle, and my understanding of the phrase “tip-of-the-tongue” is that it’s a metaphor, not a literal description of how TOTs feel. Of course, movements in the tongue muscle often accompany a TOT, given that TOT states often occur in conversation, but it does not seem that these tongue movements constitute the TOT experience.

Elsewhere Proust claims that TOT feelings have three dimensions: bodily intensity, emotional intensity, and a feeling of imminence (2013, pp. 61, 115, 135), where the strength of these feelings can guide action by prompting the subject to continue their attempts at retrieval. These claims are attached to a citation of Schwartz et al. (2000). But Schwartz et al. make no mention of TOT as embodied. They do discuss three phenomenological aspects of TOT that they probe in their experiments, describing the first as “the intensity of the TOT feeling” (2000, p. 19), but this intensity of feeling is never described as bodily. Furthermore, Schwartz et al. do not claim that an emotion is an essential component of TOT, instead, they suggest that TOTs “evoke” an emotional response like frustration, that they are often “accompanied by” emotions (2000, p. 19), and they approvingly quote influential TOT researcher Brown (1991) as explaining that such emotions “may simply reflect a reaction to the momentary retrieval failure rather than being an integral component of the TOT” (Brown, 1991, p. 205, in Schwartz et al., 2000, p. 19). I agree with Brown and the psychologists I will discuss in Sect. 6; although there may be a variety of embodied and emotional reactions to the TOT experience, they are not essential to the TOT experience itself.

A similar account of metacognitive experiences as embodied metacognitive feelings is offered by Dokic (2012). He describes “noetic feelings” according to his “Water Diviner Model” in the following way:

On this model, noetic feelings are first and foremost bodily experiences…They are diffuse affective states registering internal physiological conditions and events. Unlike bodily sensations, though, noetic feelings need not have precise locations in external bodily parts. At a phenomenological level, they often have an ‘indistinct, spreading, blurred quality’ and they ‘seem to actively resist attempts to focus attention directly on them’ (Mangan, 2001). In William James’s terms, they belong to the ‘fringe of consciousness’ (James, 1890). (p. 307)

I agree with Dokic that TOT experiences are not precisely located in any body part, but I would argue that this is a consequence of their mental phenomenality. And it does not appear that TOTs have the character of being indistinct, spreading, blurred, and resisting attempts to focus on them, instead, TOT experiences are particularly salient, grabbing our full attention when they occur. Such experiences are not peripheral or on the fringe of consciousness, rather, they are in the focus of experience, inviting our attention towards a distinct phenomenon within the mind. I am not arguing that the above descriptions are inaccurate regarding some noetic feelings, and I grant that there very well may be aspects of the TOT experience that include embodied fringe feelings of this type. But the introspective evidence I have pointed to suggests, minimally, that there is more to the TOT experience that the accounts of Proust and Dokic have left out. Compatible with the views of these authors is my position that the TOT experience includes cognitive features that constitute aspects of its phenomenal character that are not captured by the above descriptions of noetic feelings.

Dokic rejects what he terms the “Direct Access Model” of noetic feelings, and because the Cognitive view of TOT might bear some resemblance to such a model, I will show how his arguments do not apply to my view.Footnote 3 On this model, noetic feelings are instances of introspection on first-order memories. Dokic explains how this model works by giving the example of a subject having a feeling of knowing that they know the capital of Peru, even though the subject cannot produce the name (“Lima”). According to Dokic, such a subject

is conscious of her memory only as having a content of the form ‘The capital of Peru is called ____.’ In other words, she has introspective access to her memory as such while having access only to a proper part of its content. (p. 306)

The problem Dokic poses for such a case is that introspection should make the contents of phenomenal states transparent, such that all the contents are revealed (pp. 306–307). Whereas in the Lima example described above, the contents are partially occluded. In other words, normally we can identify what mental state we are experiencing based on the content of that state, but in the Lima case, we introspect that we are experiencing knowing what the capital of Peru is despite the content of that state not including the very knowledge at issue (“Lima”). As Greely (2021, p. 6807) rightly notes, solutions to this problem already exist in the psychological literature, to wit, the dominant psychological explanation of TOT, that I use as a basis for the Cognitive view, as I will discuss in Sect. 6. On this view, the phenomenal character of the TOT experience is constituted by semantic representations related to the desired word. Here, the subject does possess the relevant first-order memory, the semantic representation, such that the content of the state is transparent to the subject. Note that the phonological representation of the word is not posited as a constitutive part of the experience, so on this view, the information that has not been retrieved is rightly regarded as not being a part of the experience, therefore, no transparency problem is generated.

Arango-Muñoz (2019) explicitly situates his discussion of embodied metacognitive feelings within the cognitive phenomenology debate, seeking to undermine the position that metacognitive experiences are evidence for CP. But because Arango-Muñoz only discusses unconscious causal influences on metacognitive experiences, the evidence he presents has no clear implications either for CP or its denial.

To understand the strategy of Arango-Muñoz’s paper, it will be necessary to review how he interprets the claims of proponents of cognitive phenomenology and how he understands these to relate to metacognitive experiences. Arango-Muñoz describes the view of Goldman (1993)Footnote 4 and others following him, as interpreting TOT and other metacognitive experiences to be “clear-cut” instances of cognitive phenomenology which are “irreducible in the sense that they are nonsensory experiences of cognitive attitudes” (p. 252). The irreducibility at issue is explained by Arango-Muñoz in terms of Elijah Chudnoff’s formulation of the Irreducibility Thesis, which is: “Some cognitive states put one in phenomenal states for which no wholly sensory states suffice” (Chudnoff, 2015, p. 6 in Arango-Muñoz, 2019, p. 249). Consider how Arango-Muñoz interprets this thesis:

First, although the thesis is framed in non-causal terms, it is naturally read as a causal condition: cognitive states (thinking, remembering, reasoning, calculating) cause or produce cognitive phenomenology. Thus, if some phenomenology is to be considered as properly “cognitive” it should have a cognitive state as its cause. The second aspect has to do with the fact that, since no sensory state suffices, this type of phenomenology cannot be produced by sensory or bodily states, no matter how many or how complex they are … In the following sections, the Irreducibility Thesis will be the main criterion to assess whether some phenomenology is or is not properly cognitive. (p. 250).

Here, Arango-Muñoz interprets cognitive phenomenology to concern causal rather than constitutive contributions to phenomenal character, and so his strategy is to show that the causal antecedents of metacognitive experiences like TOT include sensory states, which Arango-Muñoz argues would undermine the view that such experiences are instances of cognitive phenomenology.

This argumentative strategy is problematic from the outset because both proponents and deniers of cognitive phenomenology understand the debate to concern constitutive and not causal contributions to phenomenal character, as explained in Sect. 2. To repeat a line quoted above, Arango-Muñoz’s interpretation of the thesis of cognitive phenomenology is that “cognitive states (thinking, remembering, reasoning, calculating) cause or produce cognitive phenomenology” (p. 250). But cognitive phenomenology concerns whether thinking, remembering, reasoning, and calculating are themselves consciously experienced, not whether (potentially unconscious) thinking, etc., cause some other type of phenomenology. So, even if Arango-Muñoz shows that a variety of sensory stimuli causally influence phenomenal character, this will not undermine the arguments of proponents, nor would it bolster the arguments of deniers.

For example, consider Arango-Muñoz’s discussion of evidence regarding the interference of bodily sensations, focusing on Goldinger and Hansen (2005), who demonstrate how participants’ judgements of familiarity are causally influenced by subliminal somatic stimuli delivered by buzzers hidden in their seats. Arango-Muñoz argues that because these results show bodily sensations influencing metacognitive feelings, the argumentative requirements of the Irreducibility Thesis have been violated (p. 255). But these results only concern causal, not constitutive, contributions to phenomenal character, therefore, while they provide interesting insights into the psychology of these states, this has no clear import for their phenomenology. These results would support the denial of CP if they showed that the phenomenal character of the “feeling of familiarity” were constituted solely by the feeling of a buzzing on one’s bottom, but note that only a subliminal, i.e. non-conscious seat-buzz influenced subjects’ judgements. In the control group where the seat buzz was strong enough that participants consciously felt it—in other words, when the sensory phenomenon at issue was part of the phenomenal character of the participants’ experience—this buzz had no effect on their judgements (Goldinger & Hansen, 2005, p. 526).

Or consider Arango-Muñoz’s discussion of metacognitive illusions and the “heuristic basis” of metacognitive feelings, which he argues demonstrates that metacognitive feelings “do not fulfil the argumentative requirement of the Irreducibility Thesis” (p. 254), where that requirement would be that metacognitive feelings “should be determined by mental attitudes themselves and by nothing else” (p. 252). Arango-Muñoz (p. 254) then points to evidence that the feeling of knowing is influenced by factors like the availability of information related to the knowledge in question, even when the subject does not possess that knowledge itself (Koriat, 1993, 2000), and that TOT is similarly influenced (Schwartz, 2002; Schwartz & Metcalfe, 2011). But even if the feeling of knowing is caused by information that is not the information that the subject feels they know, this tells us nothing about the phenomenal character of that feeling. The claim of proponents of cognitive phenomenology is not that our cognitive experiences are veridical, nor is it that our cognitive experiences must be caused by nothing other than conceptual representations, instead, their claim is that the phenomenal character of those experiences cannot be reduced to merely sensory representations. While these results present an interesting epistemic problem for metacognitive feelings, this has no clear bearing on their phenomenology.

Dorsch (2023) also argues that noetic feelings are embodied, and while he shares some of my concerns with the sort of evidence that Arango-Muñoz cites, the stronger evidence he presents is still not conclusive. Dorsch considers a range of studies showing connections between embodied states and metacognitive phenomena, recognizing that the embodied states considered by Goldinger and Hansen (2005), among others, are not (or may not) be conscious (Dorsch, 2023, pps. 11–13). The strongest evidence that Dorsch considers is from a study conducted by Fiacconi et al. (2017), where researchers investigated subjects’ interoceptive sensitivity to heartrate acceleration in relation to feeling of knowing judgements. Results showed that the more accurate subjects were at detecting heartrate acceleration, the stronger their feelings of knowing were. Dorsch argues these results suggest that “the better you are at sensing bodily afferents, the more intense will be your noetic feelings, a relationship most easily explained by how those afferents cause such feelings” (2017, 17). I grant that the heartrate acceleration at issue is conscious and that these results suggest an intimate connection between such bodily afferents and noetic feelings. And while I think it is plausible to suggest that such afferents could cause such feelings, this does not strike me as the most natural explanation. Instead, we might interpret the direction of causation to work the opposite way: noetic feelings can cause changes in heartrate. In the case of TOT experiences, which often evoke emotional responses like frustration, and which lead to the marshalling of cognitive resources to try to recall the target word, I expect that these responses to the TOT experience will involve a variety of changes in bodily arousal. But on this understanding, the TOT experience causes the embodied phenomena (heartrate acceleration), not the converse.

Ultimately, none of the authors above prove that the phenomenal character of TOT is constituted by metacognitive embodied affects, and none of the authors present a convincing case that the phenomenal character of TOT is not (in part) constituted by conceptual representations. Below, I more carefully examine their commitments and what they have or have not demonstrated regarding cognitive phenomenology.

5 Do Embodied Affect Views Exclude Cognitive Phenomenology from TOT?

According to the metacognitive embodied affect interpretation of TOT, sub-personal metacognitive faculties in the brain monitor the performance of memory. When these metacognitive monitors determine that the phonological form of a word may be present in long-term memory despite a momentary failure to retrieve it, they inform the subject by outputting to consciousness a metacognitive feeling. On this view, these metacognitive feelings are nonconceptual embodied affects, and so they are a phenomenon that deniers of CP grant can be present in phenomenal character. But do philosophers who hold this metacognitive interpretation of TOT deny that TOT phenomenology may also involve conceptual representations? Here are two possible claims that defenders of the embodied affect view of TOT could be committed to, one which would be committed to the denial of CP, and one compatible with CP:

Metacognitive Denial (MD): metacognitive feelings, understood as nonconceptual embodied affects, wholly constitute the phenomenal character of the TOT experience.

Metacognitive Compatible (MC): metacognitive feelings, understood as nonconceptual embodied affects, in part constitute the phenomenal character of the TOT experience.

According to MD, metacognitive feelings of this type exhaust the phenomenal character of TOT, whereas according to MC, metacognitive feelings are only one aspect of the phenomenal character of TOT, leaving room for conceptual representations to also in part constitute the phenomenal character of the experience.

Proust, Dokic, and Dorsch offer interesting accounts of how embodied metacognitive affects might work, but they do not explicitly discuss how their views may have import for cognitive phenomenology. Because they do not argue that TOT experience is nothing more than an embodied affect, they are not clearly committed to MD. On the other hand, they never consider how TOT may include cognitive or conceptual aspects, and so they never argue their view is compatible with such an interpretation, so they are not clearly committed to MC. None of the evidence or argumentation the authors present supports MD, so they might wish to endorse the weaker MC claim. But even the strongest evidence on offer from Dorsch is not entirely probative, leaving room to doubt whether the metacognitive phenomena identified are embodied. So, while their arguments provide some support for MC, these arguments do not seem conclusive by my lights. Their arguments would be strengthened if they could show that the embodied phenomena at issue are constitutive parts of the phenomenal character of metacognitive experiences like TOT, rather than embodied responses to or embodied co-occurences with metacognitive experiences.

Arango-Muñoz explicitly situates his argument that TOT involves metacognitive embodied affects within debates concerning CP, but because he only considers causal influences on phenomenology, it is not clear whether he is ultimately committed to MD or MC. He sets out to defeat the argument that metacognitive experiences like TOT are instances of cognitive phenomenology, and while he succeeds in showing that various sensory inputs can causally influence metacognition, he never establishes that the sensory influences at issue even partly constitute the phenomenal character of TOT, so this evidence does not provide convincing support for the weaker MC claim, let alone the stronger MD claim. Arango-Muñoz’s argument would be strengthened if he could show that these unconscious sensory causal influences necessarily produce purely sensory phenomenal character.

In sum, views that understand TOT to be characterized by metacognitive embodied affects have the potential to impact debates over cognitive phenomenology if their defenders choose to clarify how these views commit them to MD or MC. In what follows, I will present evidence that the phenomenal character of TOT is constituted in part by conceptual representations, which will undermine the case for MD. MC, then, will seem more defensible in comparison, but the claim that metacognitive experiences include embodied affects at all might also seem to be undermined by a cognitive-conceptual view of TOT, because from the point of view of the Cognitive view of TOT, it is unclear what functional or phenomenological role embodied affects have to play in TOT. That said, I expect these roles could be clarified and defended, and while I have expressed some initial resistance to embodied affects above, I do not deny they may exist, nor do I deny that they might in part constitute the phenomenal character of TOT.

The view I defend below is a cognitive interpretation of TOT, according to which the subject’s memory retrieval processes have succeeded in retrieving semantic-conceptual representations associated with the target word but failed to retrieve a phonological representation of it. The result is that these conceptual representations become a constitutive part of the phenomenal character of the subject’s experience, while the desired phonological representation is not. Although I call this a “cognitive” view, this does not imply a denial that TOT experiences may also include metacognitive aspects, I only deny that nonconceptual, embodied metacognitive affects exhaust the phenomenal character of TOT. Therefore, the claim I seek to defend is a compatible with embodied affect views, formulable as:

Cognitive Compatible (CC): conceptual representations in part constitute the phenomenal character of TOT experience.

Below, I will show how this claim is backed by substantial empirical evidence and by introspection on the phenomenology of TOT.

I should distinguish between the above compatibility claim and Calabi’s (2016) pluralist view of TOT. Calabi argues that there is not a single type of TOT experience, instead, there may be multiple types of TOTs, such that TOTs are feelings, perceptions, or second-order beliefs. By defending CC I grant that there may be a variety of phenomena that co-occur with TOT, but I argue that all TOT states share at least one phenomenal feature in common, namely, a conceptual representation of the meaning of the unrecalled word. In short, by defending CC I am arguing that all TOTs share this phenomenal feature essentially, whereas Calabi does not argue that there is any single essential phenomenal feature that all TOTs share.

6 The Cognitive View: The Psychology of Language Production and TOT

Brown and McNeill’s “The ‘Tip of the Tongue’ Phenomenon” (1966) initiates modern research on TOT. They theorized that TOTs occur when memory retrieval processes successfully access the target memory, but there is a failure to retrieve all the features of the word. To illustrate their explanation of the phenomenon, Brown and McNeill provide an analogy between long-term memory and a card catalogue system (1966, pp. 333–335). Each “card” has the name of the object, including the letters and syllables in that name, a definition, semantic features of the object, and words similar in meaning or sound. Brown and McNeill propose that during a TOT state, the subject’s memory retrieval processes have successfully retrieved the card associated with the desired word but some of the information on the card is missing or blurred. This explains the behavioral data that subjects can often report a variety of semantic and lexical features of the words they cannot recall.

TOT research continued steadily through the next twenty-five years, but two papers in 1991 had an enormous influence on TOT research. Alan S. Brown’s “A Review of the Tip-of-the-Tongue Experience” (1991) collected and distilled decades of results. Notably, Brown reviews evidence showing that partial information related to the target word is often available during a TOT state. Some of the information retrieved is semantic, with studies showing that semantically related words were reported by subjects during 50–56% of TOT experiences (p. 208). Partial phonological or orthographic information is also available: subjects’ guesses of the first letter of the target word are correct 50–71% of the time, they have less accurate guesses of other letters, and the numbers of syllables of target words can often be guessed (pp. 209–210). Reflecting on this evidence, Brown notes: “the fact that target-word information is available during a TOT strongly suggests that the target is in a state of partial activation” (p. 210). Here, Brown (1991) reaches a similar conclusion to Brown and McNeill (1966), proposing that TOTs involve memory retrieval processes making some partial information available to the subject related to the target word, even when the full phonological form remains unavailable.

The next influential paper to consider is from Burke et al. (1991), who present a new theory of TOT in the context of theories of speech production. First, it will be informative to understand the psychological models of speech production on which these explanations of TOT are based, before returning to how they explain TOT experience.

Within the psychological field of speech production, there is universal endorsement of the “two-step” model of speech production. The first step begins with semantic representations (abstract, non-sensory, non-word) of the object-to-be-named and moves to the “lemma” level where the grammatical and syntactical features of the word or phrase is formulated. The second “step” consists in constructing a phonological representation of the words-to-be-spoken. In this way, the cognitive system translates meaning to words, and words to sounds (Bock & Levelt, 1994; Burke et al., 1991; Rapp & Goldrick, 2000; Schwartz & Dell, 2015). There is a large literature discussing various specifics of these models, and there is not agreement on every nuanced detail, but overall, every current model of language production involves the activation of semantic representations prior to the activation of phonological representations (Gollan & Brown, 2006).

The evidence that supports the two-step model is drawn from a variety of well-studied linguistic phenomena, including regularities in speech errors, the time course of lexical activation, and studies of anomia, among others. There will not be time to discuss in any detail the varied literature supporting the model but let me provide a few brief examples.

Consider Garrett’s (1975) influential study of speech error patterns. He showed that word exchange errors (e.g. “that dirt has chair on it”) usually occur within different phrases and involve words of the same syntactic category. This is proposed to occur at a syntactic processing stage where elements that are computed simultaneously (e.g. nouns) might erroneously switch positions. In contrast, sound exchange errors (e.g. “Why is this dair chirty?”) usually occur within a phrase and involve words of different syntactic category. This is proposed to occur at a phonological processing stage where the elements that are computed simultaneously are those that are close to one another in serial order, such that they might erroneously switch. In relation to the Two-Step model, the first type of error happens during the first step, while the latter type of error happens during the second.

A second type of support for the model comes from experiments designed to test how speech is generated over time. Schriefers et al. (1990) directed subjects to speak the name of objects shown in pictures as quickly as possible, with experimenters interfering with this task by auditorily presenting words that were semantically related to the picture (“cat” when presented with a dog), phonologically related (“fog”), or unrelated to the picture (control). When a semantically related word was played before the appearance of the picture, reaction times were significantly longer when compared to controls, whereas when a phonologically related word was presented before, reaction times were the same as in the control condition. When a semantically related word was played at the same time or after, the interference effect disappeared. In contrast, when a phonologically related word was presented at the same time or just after the picture, a significant facilitatory effect was observed—reactions times became faster. Taken together and simplified, the results showed that semantic interference affected early speech production, while phonological interference affected later speech production. This evidence is what we should expect according to the Two-Step model, on which semantic processing occurs before phonological processing.Footnote 5

Subjects with anomia serve as another source of evidence. Consider G.M. who was tested to have normal verbal learning capacities, comprehension, memory, and vocabulary, but a pronounced difficulty with naming, with a roughly 50% failure rate in picture naming tasks. Most striking is what G.M. could say about the objects, even when he could not name them. He would often state that he knew what the object was, and he would often demonstrate his semantic knowledge of the object (reporting what it’s used for, similar objects, etc.). He could often guess lexical features of the name including its gender, approximate length, and its first letter. For example, when shown a picture of a bird cage (une volière) he said: “… it starts with V … they can fly … it starts with VOI … in museums … when birds can fly inside … it’s feminine, I am sure of it … it sounds like sailing boat (voilier)…” (Henaff Gonon et al., 1989, pp. 395–396). G.M.’s speech capacities and deficiencies are easily explained on a Two-Step model of language production; he often remains stuck on the first step, with access to semantic and lemma-level features, but his language system fails to complete the second step of retrieving the word’s phonology.Footnote 6

These and a variety of other well-studied phenomena combine to support the two-step model of language production. Now let’s turn to how such models explain tip-of-the-tongue experience.

Burke et al. (1991) argue that TOT states occur when there is successful retrieval of semantic representations and possibly lemma-level features (i.e., some of the first step has completed) but there is a failure to retrieve a full phonological representation of the word (i.e., the second step fails to complete). This explanation was widely accepted at the time and remains dominant today among psychologists who research TOT (Bock & Levelt, 1994; Brown, 1991; Caramazza & Miozzo, 1997; Gollan & Brown, 2006; Vigliocco et al., 1999, for recent studies that explain their results on this model see Bannon et al., 2024; Chang et al., 2022; Cleary et al., 2021, Dewi & Salikin, 2022, Huebert et al., 2023). Reviewing any further nuanced information regarding this view would not be profitable for our purposes. Although it may be instructive to read a concise explanation of TOT in the words of psychologists:

When a TOT occurs, a lexical node in the semantic system becomes activated, giving access to semantic information about the target word, but at least some phonological information remains inaccessible because insufficient priming is transmitted to enable activation of connected phonological nodes. (Burke et al., 1991, p. 545)

the TOT phenomenon is a failure to access the lexeme from the lemma. The speaker knows the meaning to be expressed (i.e., the concept) and the word’s syntax (that it is a plural noun, a transitive verb or whatever; i.e., the lemma). Only the word form is blocked. (Bock & Levelt, 1994, p. 953)

According to these psychologists, tip-of-the-tongue states occur when semantic-conceptual representations, the meaning of the word, is successfully retrieved, but the phonological form of the word is not.

Before turning to what this implies for cognitive phenomenology, it might be helpful to explain how to interpret this talk of “access” and “activation.” Psychologists consider two-step models to show the functional organization of neural networks. Activation in the network spreads from semantic nodes to lemma nodes, and finally to phonological nodes. When a particular node reaches some threshold of activation, the contents of that node can then be accessed, i.e., become conscious: “activation of nodes is necessary to consciously retrieve information a node represents” (Burke et al., 1991, p. 543), and Caramazza and Miozzo (1997) also explicitly equate access with “conscious retrieval” (p. 333). Psychologists understand TOT, which occurs when conceptual representations are accessed, to be a phenomenally conscious experience, explicitly referring to TOT “phenomenology” (Schwartz, 2002; Schwartz, 2002, Cleary et al., 2021; Huebert et al., 2023), TOT as a “subjective” state (Brown, 2008; Caramazza & Miozzo, 1997; Huebert et al., 2023; Taylor & McKay 2003), or TOT as an “experience” (Brown, 2008; Gollan & Brown, 2006; Bannon et al., 2024; Dewi & Salikin, 2022; Huebert et al., 2023). Note that introspection on the phenomenal character of TOT is crucial to psychologists: the experimental subject is commonly recorded as being in a TOT state if and only if they experience themselves as seeming to be in one. Even when a subject feels they are in a TOT state despite not having the target in memory, many psychologists label these as inaccurate or illusory TOT states, rather than non-TOT states, because they define TOT states subjectively. In short, the psychologists discussed above are seeking to explain a subjective, phenomenally conscious TOT experience, which is detected by introspection on the phenomenal character of that experience, by appealing to the underlying mechanisms of memory retrieval and language production that give rise to it. And specifically, what gives rise to that experience, according to these psychologists, is access to semantic-conceptual representations, in the absence of the desired phonological representation.

7 Heuristic-Metacognitive View of TOT: The Psychology of Metacognition

While the above view of TOT based on two-step models of language production is a long-standing and dominant view of TOT, more recent views inspired by models in the psychology of metacognition present a contrasting alternative. My view is that these models are neutral regarding their import for debates concerning cognitive phenomenology, but it will be worthwhile to discuss them briefly to show how they could be modified to be relevant to these debates.

Schwartz and Metcalfe (2011) present a view of TOT based on models of metacognitionFootnote 7. On this understanding, the “TOT feeling” is produced by a metacognitive monitoring mechanism that examines a variety of information related to the target word. The (non-conscious) input to this monitoring mechanism includes familiarity with the target, memories of past recall of the target, and the semantic, phonological, or syntactic information anticipated by two-step models. The output of this monitoring mechanism is a TOT feeling that is phenomenally conscious, serving to inform the subject of the performance of unconscious memory retrieval processes. Note that Schwartz and Metcalfe make no claims regarding the representational format of the conscious TOT feeling that is the output of the monitor, with no suggestion that the phenomenal character of such a feeling is either semantic-conceptual or sensory-embodied. The motivation for this model includes experimental results that showed interactions between TOT states and metacognitive phenomena. For example, Cleary (2006) showed that when subjects were in a TOT state, they were more likely to judge that the unrecalled target word was presented in an earlier trial (a recency or familiarity effect), even though TOT states did not correlate with whether the target was presented earlier. More recently, Cleary (2019) demonstrated an association between TOT and positive valence (a “warm glow” effect), with results showing reports of TOT were more likely for target words whose questions suggested the targets were positively valenced, and that subjects in TOT states were more likely to infer positive characteristics regarding the unrecalled word. While these results are compatible with transmission deficit models and the Cognitive view of TOTFootnote 8, these interactions with metacognitive phenomena have been interpreted to suggest that access to semantics is not the only variable that affects TOT.

Even if these results were strictly incompatible with the Cognitive view, metacognitive models like Schwartz and Metcalfe’s (2011) have not been defended in a way that is relevant to the affirmation or denial of cognitive phenomenology. I’ve already made this argument in relation to Arango-Muñoz (2019) in Sect. 4, but it is worthwhile to restate this argument now that I have reviewed the metacognitive model of TOT, and because I think Arango-Muñoz’s approach to these issues may be shared by other philosophers who endorse metacognitive embodied views. First, let me repeat that Schwartz and Metcalfe (2011) make no claims regarding the representational format of the TOT feeling, so on their model it seems plausible that the TOT feeling could be a semantic-conceptual representation, a sensory-embodied representation, or a sui generis experience that could challenge the sensory vs. conceptual dichotomy at play in cognitive phenomenology debates. Recall also that while the cognitive phenomenology thesis is commonly understood to concern the constitution of the phenomenal character of conscious states, Arango-Muñoz interprets it as a causal thesis, claiming that for a state to be an instance of cognitive phenomenology, it should be caused by only conceptual information. Presumably a parallel claim should hold for those who defend sensory-embodied views: for a state to be an instance of sensory phenomenology, it should be caused by only sensory-embodied information. To accommodate the sort of evidence Arango-Muñoz appeals to, let’s update the Schwartz and Metcalfe (2011) model to include a variety of sensory-embodied inputs to the metacognitive monitoring mechanism. The most fundamental problem for the embodied metacognitive thesis is that there is no necessary connection between the unconscious input to the metacognitive monitoring mechanism and the conscious output. According to the Cognitive view, access to semantics makes those conceptual-semantic representations conscious, without the positing of some intervening mechanism that takes those representations as both input and output. Metacognitive monitoring views do posit an extra step, a mechanism that intervenes between unconscious inputs and the phenomenally conscious state that is the output of that mechanism. For these metacognitive monitoring views to warrant claims relevant to cognitive phenomenology, Arango-Muñoz and his peers need to provide evidence and argumentation concerning the output of those mechanisms, but instead, they only provide evidence concerning the input. Furthermore, notice that the input to these mechanisms is mixed regarding representational format; Schwartz and Metcalfe (2011) argue that these inputs include semantic-conceptual representations, as well as non-sensory but non-conceptual syntactical representations, and in line with embodied views, let’s grant that a whole variety of embodied and sensory information also serves as input to the monitoring mechanism. The output, the conscious TOT feeling, would then count neither as cognitive phenomenology, nor as sensory-embodied phenomenology, according to Arango-Muñoz’s cognitive criteria and the parallel sensory formulation, because the output is causally connected to input that is neither purely cognitive nor purely sensory. In short, the psychologists defending metacognitive views of TOT do not characterize their own models in a way that is relevant to cognitive phenomenology, and the philosophers who appeal to these models to try to defend an embodied interpretation of the TOT experience do not properly justify their claims.

Aside from the philosophical clarifications that I have suggested would strengthen arguments for embodied metacognitive views, there are also ways that psychologists could productively move this debate forward. For example, Huebert et al. (2023) showed that subjects in TOT states are often inaccurate in their reporting of phonological or syntactic features of unrecalled target words, which provides evidence against Direct Access views of TOT. If an alternative experimental design could demonstrate that subjects in TOT states are often inaccurate in their reporting of semantic features of unrecalled words, this would provide evidence against the Cognitive view of TOT. It would also be informative to investigate the occurrence of TOT (or lack thereof) in subjects with disordered interoception. If the phenomenal character of TOT is constituted by embodied phenomena, then we can predict that those who have impaired conscious awareness of their own bodies should have fewer or weaker TOTs than subjects with normal interoception. In addition to these behavioral investigations, psychologists could ask subjects to introspect on their TOT states, for example, whether they are experienced as located in the body or mind, or whether subjects experience the meaning of the word in the absence of its sound as opposed to experiencing a more diffuse affective state.

8 TOT and Cognitive Phenomenology

In this section I return to the Cognitive view of TOT to explain precisely how it relates to cognitive phenomenology and to reply to some objections. Many psychologists understand TOT experiences to occur when semantic representations related to the target word are retrieved, but the full phonological representation is not. Semantic representations are abstract, non-sensory representations of the meaning of words, and so they are conceptual representations. Note that such representations are in no way composed of words or sounds, instead, they are accessed in the first stage of language production before access to syntactical, grammatical, or phonological features. Recall that the first step in the model begins with access to semantic-conceptual representations and completes with access to lemma-level features, while the second step proceeds from a lemma to the retrieval of phonological representations. The Cognitive view of TOT posits that in a TOT state, the conceptual representations that are retrieved at the beginning of this process will always be part of the phenomenal character of a subject’s TOT experience, sometimes the lemma-level features retrieved in the middle of the process will also be part of the phenomenal character of a subject’s experience, and less frequently the phenomenal character of a subject’s experience will also be in part constituted by some partial phonological features retrieved towards the end of this process. This explanation accords with earlier research from Brown and McNeill (1966) and Brown (1991), recognizing that partial information related to the target word is often available, but this explanation specifies which information is available, when, why, and how. To wit, semantic representations are available, they are available before any potential partial phonological representations, this is because semantic retrieval precedes phonological retrieval in the speech production process, and they are available insofar as they are introspectively accessible constitutive parts of the phenomenal character of the TOT experience. This is why a subject in a TOT state experiences knowing what they mean, despite the inability to say it, or knowing who it is even when they cannot be named. This view draws attention to the cognitive aspects of TOT, the subject’s experience of a conscious first-order mental state constituted by a semantic-conceptual representation, but it remains neutral on potential metacognitive aspects of TOT, i.e., this view is committed to CC.

Finally, consider how CP follows from the Cognitive view of TOT. Because this view explains the TOT experience in terms of the phenomenal character of a subject’s experience being constituted by conceptual representations, where to show that CP is true, the proponent needs to identify a conscious experience whose phenomenal character is constituted, at least in part, by conceptual representations. TOT fulfills this condition, therefore, CP. These conceptual representations are semantic representations, which are experienced as the meaning of a word or understanding who it is, despite the absence of the sound of the word or name, which are robustly cognitive phenomena that all would agree are sufficient to demonstrate CP.

Do the psychologists whose views comport with mine have a stated position regarding whether TOT includes cognitive phenomenology? In other words, do these psychologists understand access to semantic representations to constitute or merely cause the TOT experience? Note that these scientists do not explicitly commit themselves to either position, because they do not use such precise philosophical terminology, but a constitutive reading is the most reasonable to infer from the evidence and interpretations of it that they provide. From the passages quoted above, these psychologists understand TOT to occur when semantic representations are accessed in the absence of access to phonological representations, where psychologists understand access to imply that the representations accessed are conscious, and understand TOT to be a phenomenological experience. Psychologists are aiming to explain this phenomenological experience by appealing to what occurs during the two-step process of language production, where this explanation is taken to be sufficient to explain the phenomenological experience of TOT. If the semantic representations accessed do not constitute the phenomenal character of TOT, and there is some other phenomenon that constitutes the phenomenal character of TOT, presumably, these psychologists would address this extra step that intervenes between access to semantic representations and the experience of TOT. A causal interpretation compatible with MD would require the positing of such an extra step, but these psychologists do not posit any such intervening step. Therefore, it is reasonable to infer that these psychologists understand the phenomenal character of TOT to be constituted by the semantic representations accessed, rather than merely being caused by these semantic representations where the experience is constituted by some other type of representation that they never mention.

The denier of CP might object to this constitutive reading of the psychological data concerning language production and argue that it is some sensory or embodied phenomenon that constitutes the phenomenal character of TOT. But the denier is then committed to positing an extra step that is missing from the cognitive explanation of TOT. If we can explain both the psychological function of TOT and the phenomenal character of TOT according to the Cognitive view, then this extra step appears unnecessary. Consider that Greely (2021) has also argued that the sort of metacognitive views I examine in Sect. 4 are unmotivated, because there are viable explanations of metacognitive experiences in terms of first-order mental states, comporting with the Cognitive view. So, the denier is free to argue for a causal reading of the psychological data, and to posit some additional step by which access to semantic representations causes the production of some nonconceptual, sensory or embodied metacognitive experience, but they would need to justify why we should posit such an additional step, when the Cognitive view fully explains the phenomenon at issue in a more parsimonious way. Relatedly, the denier might posit that the additional step is the production of an embodied affect, for example, a feeling of frustration, where this emotional response is what we expect regularly occurs in relation to TOT – therefore, it doesn’t seem like such an “extra step.” But, technically, this is an extra step from the point of view of the psychological models of TOT reviewed above. Furthermore, recall from Sect. 4 that some psychologists explicitly state that such affects are responses to the TOT experience, rather than a constitutive part of the TOT experience. Here, the experience of TOT is understood to precede the affective response to it. So, the denier would need to provide an argument for why we should consider these affects as constituting the TOT experience, rather than being responses to it.

Another objection from the denier of CP might be to accept that my cognitive explanation of TOT is accurate, but to argue that the conceptual representations accessed are insufficient to demonstrate CP. For example, one might think that the conceptual representation at issue is a description of the referent of the name that one cannot recall, for example “the lead female android in the movie Blade Runner” where this description is expressed in inner speech, such that the phenomenal character of one’s experience during TOT is constituted by this sentence in inner speech. Because this conceptual representation takes the form of a sentence in inner speech, it is insufficient to demonstrate CP. Such an objection would misread both the psychological literature and the commitments of proponents and deniers of CP. As explained above, the conceptual representations at issue are semantic representations, which psychologists of language production understand to be non-sensory and non-phonological, so they are not representations that take the form of inner speech sounds, and they are not representations by way of words. Consider also that even if one does experience some sentence in inner speech, according to the denier, this sentence in inner speech only constitutes the phenomenal character of a subject’s experience insofar as that sentence is a purely phonological representation, and any conceptual representations that may have been involved in the generation of that phonological representation never constitutes the phenomenal character of a subject’s experience (see e.g. my discussion of Tye in Sect. 2 above, or Prinz, 2012, esp. pp 69, 150–153). So, any description entertained in inner speech that constitutes the phenomenal character of one’s experience is not a conceptual representation, according to deniers of CP. So, it is not open to the denier to object that the conceptual representation that constitutes the phenomenal character of experience according to the Cognitive view is in fact merely a sentence in inner speech, because according to the denier, no sentence in inner speech is a conceptual representation insofar as it constitutively contributes to the phenomenal character of one’s experience.

It should also be noted that the experience of a description in inner speech plus some affect is not sufficient to explain the phenomenal character of TOT. Consider that I can experience a description in inner speech like “the name of the capital city of Madagascar” and feel frustrated that I cannot produce the name and not feel that I am in a TOT state. In fact, I have never known anything about the capital of Madagascar, including its name or semantic features, and so I don’t experience that the name is on the tip of my tongue. But if it were true that the phenomenal character of TOT is constituted by nothing other than sentences in inner speech plus an affective responsive, then just experiencing those sentences and that affect would be sufficient to experience TOT, but this is plainly false, because we can experience these sensory states without experiencing a TOT. The phenomenal character of TOT includes something more than these sensory states, and the Cognitive view provides an explanation for what that something more is.

Finally, I should clarify the difference between the experiences of conceptual understanding we experience in TOT according to the Cognitive view, and experiences of linguistic understanding as feelings of fluency (Drożdżowicz, 2021). Drożdżowicz discusses experiences of understanding the meanings of words, where she explains such experiences as including feelings of fluency that arise from metacognitive monitoring processes. One might think that feelings of fluency and experiences of meaning are both involved in understanding, but these experiences are distinct: a feeling of fluency is a metacognitive experience that a word has been understood whereas an experience of the meaning of a word is a cognitive experience of an understanding of the word’s meaning. This difference can be illustrated using two contrast cases. A feeling of fluent understanding is experienced by a monolinguistic English speaker when they hear the sentence “this curry is too hot,” but this feeling is not experienced when the same sentence is spoken in an unknown language. Now consider an English speaker experiencing the meaning of the sentence “this curry is too hot” where “hot” is interpreted as referring to the temperature of the curry, and then a moment later experiencing a different meaning when they interpret “hot” as referring to spiciness. An experience of fluency cannot explain the phenomenal difference between this interpretative switch case because such a feeling would remain the same in both instances, but an experience of conceptual meaning does explain the difference in one’s phenomenal character before and after the interpretative switch. In recognition of this, Drożdżowicz explicitly clarifies that her view is not meant to explain contrast cases of this sort (6, 2021).

9 Conclusion

In this paper I have argued for a Cognitive view of TOT, which holds that the phenomenal character of the TOT experience is in part constituted by conceptual representations, from which CP follows.

By arguing that the Cognitive view of TOT supports CP, I have not argued that the phenomenal character of the TOT experience is solely constituted by conceptual representations. Rather, I have argued for CC, so it is compatible with my view that TOT is also in part constituted by non-conceptual representations. For the denier to show that TOT does not support CP, the denier must defend MD and show that the phenomenal character of the TOT experience is exhausted by, is nothing more than, or is solely constituted by non-conceptual representations, as by embodied affects of some sort, where no part of the phenomenal character of TOT is constituted by conceptual representations.

The psychological explanation of TOT that forms the basis of the Cognitive view of TOT is long-standing and well supported in the psychological sciences. TOT has been understood by these psychologists to be a phenomenally conscious experience that occurs when a subject experiences semantic-conceptual representations related to a desired word when the phonological representation cannot be retrieved. Even those psychologists who defend metacognitive views never claim that the phenomenal character of TOT is constituted, at least in part, by semantic-conceptual representations, at least as far as I am aware. So, comparatively, while there is tremendous empirical support for the Cognitive view of TOT, there is no such support for an embodied affect view of TOT that excludes TOT’s cognitive aspects (MD). To defend such a view would require defending MD and responding to the challenges outlined in Sect. 5.

The Cognitive view of TOT is also in accord with introspection on our TOT experiences. When a person’s name is on the tip of your tongue you experience knowing who it is without recalling their name, and when you forget a word, you experience knowing what it means despite being unable to say it. You may experience accompanying embodied affects in response, but it would be inaccurate to describe your experience as nothing more than such an affect, denying the conceptual understanding that is central to the phenomenal character of this experience. These experiences are exactly what we should expect given the psychological models of TOT based on the science of language production, so here we find that the psychology and phenomenology of TOT are in a mutually confirming agreement.

It remains an open question to what extent the phenomenal character of TOT includes metacognitive feelings of some type. My defense of CC in this paper has remained neutral on this question, neither affirming nor denying that embodied affects or other metacognitive phenomena may be part of TOT phenomenology. This is because the proponent of CP is not committed to denying that we experience any number of sensory states, they are only committed to affirming that we experience conceptual ones. In contrast, the position of the denier of CP, or MD, is more extreme, because they are committed to denying conceptual experience, and so the burden of proof is on them to substantiate their view, especially when confronted with the phenomenological and psychological evidence presented above. Given these challenges, MC may appear a more defensible position, and if there are essential aspects of TOT phenomenology that cannot be explained by conceptual representations alone, this would open the way for a role for metacognition. But given the challenges I have posed in this paper, competing cognitive and metacognitive interpretations need to be considered carefully, and the role of metacognitive phenomena, both at the psychological and phenomenological level, needs to be explained. Moreover, defenders of metacognitive feelings should consider three possible views: they are nonconceptual embodied affects, they are conceptual representations, or they are sui generis experiences. Assuming any one of these interpretations from the outset is unjustified.