Davidson’s standpoint, from which his thesis makes sense and is plausible, begins from his assumption that “radically different representation schemes” (Roskies this collection, p. 2) govern in animals and humans.[4] However, such an assumption clearly opens up a plethora of new issues. Roskies targets these by drawing attention to the empirical concerns mentioned above, thereby showing that an empirical foundation to support Davidson’s background assumption is missing.
To my mind, these further issues resulting from Davidson’s background assumption are the following:
1) How can we defend the intuition that animal and human cognition differ in kind? In order to defend this view, it would seem that one needs to identify a distinguishing criterion that can account for the diverging representation schemes. It also has to be shown that this factor is responsible for abilities that one group of cognizers has and that is at the same time missing in the other group. If language possession were to count as such a factor, it remains to be shown which abilities hinge on its possession and execution. At the same time, following this approach, it apparently needs to be established that no non-linguistic creature cannot execute a similar ability, not even in a partial or proto-form. This difficulty leads us to issue 2:
2) How can we understand representation schemes in animals if we do not suppose a kinship to our own cognition? As Roskies rightly states, we cannot but credit animals with numerous cognitive abilities, given their at times complex and often obviously intelligent behavior. Interpreting this behavior without acknowledging any dependence on sensory states, memory, and certain motor skills, affects, and even social competencies, seems impossible. The representation schemes employed crucially depend on physiological implementation. If the physiological basis for the acquisition of environmental information is alike in humans and animals, how different can the representation of environmental information in terms of sensory and affective representations be? Even if a complete overlap between human and animal perception cannot be argued for on the basis of isomorphisms, we can (and perhaps must) commit ourselves to the systematicity of behavioral cause-and-effect relations. It is this systematicity that leaves little room for interpreting animal cognition (at least in the sensory and affective domain) as being radically different from ours.
3) In light of Davidson’s interpretationism, how much weight does language possession carry in terms of our ability to interpret other cognitive agents? When we think of how we “make sense” of another person, we rarely rely exclusively on the other’s verbal utterances. Rather, it would seem that we generally seek to compare the contents of their verbal utterances with their overt behavior; we hold another responsible, as a rational agent, if her expressed intentions diverge “too much” from her behavior. Think of the following case: your neighbor tells you about his plans to save some money for the upcoming summer vacation; the next day you see him walk into the local casino where you know he spends quite some time—and usually loses a fair amount of money. In this case, we would probably be inclined to disregard the verbal utterance (“I’m saving up for a nice summer vacation overseas”), and rather take his actions (which might involve compulsion or gambling addiction) as indicators of his real motivations and driving forces.
4) Considering this case, we can ask which role the analysis of another’s beliefs play in interpreting and whether verbal utterances are a true mirror of internal thought mechanisms and proper beliefs.
5) To my mind, the most salient question is whether we can understand human cognition, especially thought, with the help of notions like beliefs (regardless of whether they are faithfully uttered or not) and their conveyance via language. Since the discussion of this issue will require some space, I shall dedicate a proper section to it below.
We can understand why Davidson (and with him many others)[5] posits the possession of language as a necessary condition for having propositional attitudes. Namely, one may come to the view that the way a human cognizer experiences her thoughts is predominantly conveyed by her sense of inner speech.[6] Consider for a moment what it feels like to think.[7]
Probably the most prominent, identifiable feeling related to thinking is that of your inner voice, commenting on the world around you and the world inside you, making you feel distinct from, yet embedded within it. Let’s call this phenomenon—if you can follow me here—the inner-speech view[8] with regard to thinking. I will argue that this view is misleading. Our intuitive description of what the inner-speech view comes down to is intricately linked to our ability to express the contents of our thoughts in words—the form of thoughts are, presumably, sentences that are composed of concepts and words, in our minds.
But is this identification of thought with mental speech justified? For Vygotsky (1934/1987), it is clear that there are large parts of thinking that do not rely on verbal expression: “There is a large range of thinking that has no direct relationship to verbal thinking” (Vygotsky 1934/1987, p. 115). Such a view thus allows for other, non-verbal types of thought, such as pictorial or imagistic ones, such as come to bear, for example, in mental-rotation tasks or mental imagery (Shepard & Metzler 1971; Weiskrantz 1988; Kosslyn et al. 2006).
If these instances can be found, and identified as kinds of thinking, the hypothesis that language is the one and only tool for producing thoughts in us seems simply false. That thought is exclusively verbal appears thus as a form of theory-induced illusion. One might say that the fixation on language prompted by the analytic tradition has thus resulted in the projection of the method (the analysis of language) onto the phenomenon (the human mind).
Contemporary philosophy of mind left the method of linguistic analysis behind some time ago, and in order to get away from the language-bias we should shift our focus from the surface structure of thinking, namely its intentional and phenomenological (inner-speech) characteristics, to the sub-personal level of the underlying mechanisms and production schemes of thinking.
Such a reductive approach is already in place in the numerous research efforts in cognitive science that aim at describing and explaining information processing in the brain: sensory and affective components of cognition, as well as aspects of motor behavior and memory are studied in a very promising way—in the animal as well as the human domain. The problem is that our faculty of “thinking” is in this research program a rather elusive phenomenon, for various reasons: unlike when studying the neural basis of perception, for example, thought processes cannot be studied on a cellular level, since the identification of a stimulus is virtually impossible: in vision, a stimulus is light hitting the retina, whereas the “stuff” of thought is information provided by the stimulus-processing areas, thus, an “inner-system” medium. Localizing brain areas involved in thought and thinking, on the other hand, is possible. The prefrontal cortex has been shown to be involved in planning future actions and other high-level cognitive tasks (Goldman-Rakic 1996; Fuster 2008); however, this structure is strongly connected to a wide network of other cortical areas and imaging studies show that high-level cognitive tasks often if not always result from correlated activity in multiple areas across the whole of the cortex (Fuster 2008) which makes the individuation of the “center of thought” rather difficult.
In light of these complications, it is helpful to highlight the function that higher-cognitive abilities have with regard to our overall behavior. Most researchers and philosophers would agree that what this involves is the conscious representation of objects, including the deliberate manipulation of information, retrieved from memory as well as from present and actual stimuli, for the purpose of problem solving, decision making, social interaction, communication, and action planning. The involvement of language-processing areas in the execution of these tasks has already been shown (see e.g., Goel et al. 2000)—but does this suffice to support the claim that language is a necessary cornerstone of the neural basis of higher-level cognition in humans?
When “thinking” is divided and described in terms of its functional rather than phenomenal properties, the question of how far thinking relies on our capacity to speak or use language can be replaced by the question of which brain areas and input–output relations we find involved in the faculties mentioned above. This program requires a reorientation in terms of research methods and a redefinition of the phenomenon: the phenomenological description of “thinking”, e.g., in terms of inner speech, does not supply us with an understanding of its underlying processes and mechanisms. It is these, however, that we should know first before we can put our finger on the role that language (the inner and external version alike) plays in the execution and the production of the cognitive capacities listed above.
When we cannot help but attribute the ability to manipulate information in a creative way to animals and intuitively call this “thinking” (think of the Kea, a species of bird known for its curiosity and astonishing abilities in handling difficult mechanisms—they can virtually break into a safe; cf. Auersperg et al. 2009; Huber & Gajdon 2006; Werdenich & Huber 2006) we seem to have found a satisfactory criterion for crediting animals with a form of demanding cognition, not unlike our own, even though we cannot claim to understand what it feels like or how the world represent itself to the Kea.
Such a language-independent form of high-level cognition might rule in us as well, such that it precedes the formation of beliefs we form on states of the world and their linguistic representation. It might be the case, and this is the point I want to stress in this commentary, that we fall in a systematic way for a fallacy of experienced thinking, which presents us with a linguistic representation of the contents of thought, whereas the mechanisms producing these thoughts may not rely and are not caused by speech and language involving neural mechanisms.
One can object that this is not what Davidson had in mind when he claimed that thought depends on language. Davidson’s idea rests (so goes the defense) upon the assumption that language is a universal format of information processing unique to humans (in the first place) and an instance of cognition, which lies at the core of human cognition, regardless of its temporal and causal involvement in the production of thoughts. But this—so I want to claim—amounts to a phenomenological argument, even if Davidson presents it as a theoretical one. So even if language were the universal format of human thought, the empirical basis for such a claim would be quite opaque, and any theoretical argument so far rests on this weak empirical basis.