Does language play a constitutive role in human thinking? Can we literally think in language, using sentences of a natural language as vehicles of thought? I argue that we can, though not in the way often supposed.
The standard view holds that language-based thought involves dedicated neural mechanisms, which manipulate representations of natural language sentences. By contrast, I propose that language-based thinking is an intentional activity, something we do with language, like informing or arguing. Like those activities, it depends on brain processes, but its character as a form of reasoning is evident only at the personal level.
Inner speech and conscious reasoning
My account centres on how language enables the formation of consciously held epistemic and conative commitments. When you make up your mind about something, you are not passively acquiring a new belief or desire. You are committing yourself to treating a proposition as a premise or goal in your conscious reasoning, working out what it entails, and regulating your decisions and actions accordingly.
These commitments are typically formed and manipulated using natural language. We articulate premises to ourselves in inner speech, construct arguments by linking sentences together, and evaluate hypotheses by rehearsing them linguistically. Language provides the material with which active reasoning is constructed.
Inner speech plays a central role in active reasoning. Following Peter Carruthers, I argue that inner speech involves the mental rehearsal of utterances; we initiate the motor commands for speaking while suppressing the actual movements, generating sensory representations (auditory images) that can be targeted by attention and globally broadcast. But inner speech is not just a format for thinking; it is an activity, whose functions are continuous with those of outer speech. This continuity thesis accords with developmental evidence: children first use private outer speech to guide their behaviour (imitating the verbal guidance they receive from adults) and then internalize this as inner speech.
The cyclical model
Intentional reasoning follows a cyclical pattern: produce an utterance (or image of one), perceive it, interpret it as posing a subproblem, let autonomous processes generate a response, articulate that response, repeat. The intentional actions involved break down the initial problem into simpler subproblems that can be solved by autonomous processes. I describe the process as one of deliberative mastication—chewing it over, as we say.
For example, in deciding whether to attend a party, I might start by articulating a question (‘Do I really want to go?’), which evokes an automatic response (‘Henry will probably be there’), which poses a further subproblem (‘What will Henry say?’), and so on, until I reach a conclusion. The process is formally similar to doing long division: breaking down a complex problem into simpler steps, each of which I can solve intuitively.
The judgements and decisions that terminate a reasoning process are also linguistic acts. When I conclude my reasoning about the party by saying to myself ‘I won’t go’, I am not expressing a pre-existing decision. I am making a commitment to a course of action—committing myself to the plan of not going and to regulating my future reasoning accordingly. These self-directed commitments constitute distinctively conscious attitudes, which may differ from non-conscious attitudes formed by automatic subpersonal processes. They play a stabilizing role in our mental lives, giving us self-awareness and increased self-control.
Language origins and the invention of invention
This language-involving cognitive activity constitutes a distinct level of mentality—what I call the supermind—which is realized in, but functionally distinct from, the basic mind we share with other animals. The supermind is a virtual system, created through culturally transmitted habits of thought rather than dedicated neural architecture.
The supermind’s emergence required two developments. First, as Daniel Dor has argued, language had to evolve from a system for sharing immediate experience into one for instructing the imagination—for allowing humans to communicate about things distant in time and space. Second, humans had to discover how to use this capacity privately, applying to themselves the strategies of questioning and prompting they first developed for social interaction.
This explains how language-based cognition could have evolved so quickly (it did so through cultural learning rather than neural rewiring), why System 2 reasoning has the features it does (they are all features of intentional action), and how humans became genuinely creative and innovative. Through language, we created new minds for ourselves: conscious, reflective, capable of sustained deliberation. We invented the very capacity for invention.
Key publications
Natural language and virtual belief (1998) — I argue that some beliefs—virtual beliefs—are actively formed by adopting premising policies, and that forming them typically requires language. Virtual beliefs are realized in basic beliefs about one’s cognitive commitments. Language provides the scaffolding for articulating, manipulating, and evaluating these commitments. In P. Carruthers and J. Boucher (eds.), Language and Thought. Cambridge University Press.
Evolving the linguistic mind (2010) — I examine how language-based cognition could have evolved, arguing that the linguistic mind is a virtual system created through culturally transmitted habits rather than new neural structures. The supermind emerged when humans privatized and internalized social practices of argumentation.
Inner speech and outer thought (2018) — I reconcile two apparently conflicting views: that inner speech provides a format for conscious thought, and that it’s an activity continuous with outer speech. Conscious thinking is an intentional activity that follows a cyclical pattern—produce an utterance, interpret it as posing a subproblem, generate a response, articulate it, repeat. This activity can be conducted overtly in private speech and is then internalized as inner speech. Conscious judgements and decisions are self-directed commitments expressed in speech.
Popular writing
Our greatest invention was the invention of invention itself (2022) — Drawing on Daniel Dor’s account of language origins and Dennett’s theory of consciousness, I argue that hypothetical thinking emerged through cultural evolution. Language enabled humans to instruct each other’s imaginations, and inner speech privatized this process, creating the conscious mind as a virtual system. This capacity for mental self-stimulation was the invention that made sustained innovation possible. Psyche.
Related topics
The two-level mind—Virtual beliefs and premising policies
Dual-process theories—System 2 as self-stimulation
Illusionism—The virtual mind and consciousness
