This book presents a new theory of grammatical categories - the Universal Spine Hypothesis - and reinforces generative notions of Universal Grammar while accommodating insights from linguistic typology.
AUTHOR: Jackendoff, Ray TITLE: Language, Consciousness, Culture SUBTITLE: Essays on Mental Structure SERIES: Jean Nicod Lectures PUBLISHER: MIT Press YEAR: 2007
Mathias Schenner, Center for General Linguistics, Typology and Universals Research (ZAS), Berlin
SUMMARY The book aims to give an explicit account of the mental structures involved in a variety of cognitive domains: language, consciousness, complex action, theory of mind, and social/cultural cognition. It consists of two parts. Part I, an augmented version of the Jean Nicod Lectures in Cognitive Philosophy that Ray Jackendoff gave in 2003, contains a more general exploration of language, consciousness, action, social interaction and culture from a cognitive point of view. Part II focuses on social cognition and theory of mind (and their expression in language) and develops more formal analyses of crucial concepts in these domains within the framework of Conceptual Semantics.
Chapter 1, ''Mental Structure'', emphasizes the importance of studying the functional organization of the human mind for understanding how the mind/brain works, especially in view of the still limited methods of neuroscience. A sketch of the phonological, syntactic and semantic analysis of the simple sentence ''The little star's beside a big star'' serves to illustrate the complexity of mental structures. (The term ''mental structure'' is used instead of ''mental representation'' in order to avoid the impression that mental elements need to be intentional or ''about'' something mind-external.) The following questions for the study of mental functioning are identified: (a) What are the levels of mental structure and what interfaces are there? Which levels of structure are hierarchical, and which are recursive? (For example, syllabic structure is hierarchical but not recursive, syntax is both.) (b) How are mental structures processed over time? (c) How are they learned?
Chapter 2, ''Reintegrating Generative Grammar'' (based on Jackendoff (2002)), tries to explain why linguistics and generative grammar were taken to provide the key to understanding the mind in the 1960s, but have become isolated from the other cognitive sciences in the following decades. What has been right about generative grammar from the start are its three founding themes in Chomsky (1965): (1) Mentalism: The appropriate object of study is the linguistic system in the mind/brain of the individual speaker. (2) Combinatoriality: Speakers have rules or mental grammars as part of their unconscious linguistic knowledge. (3) Acquisition: There is an unlearned component, Universal Grammar, that shapes the acquisition of language. One of the reasons why linguistics lost prestige was the broken promise that the syntactic level of Deep Structure would be the key to the mind. In addition, Jackendoff points out two ''scientific mistakes'' in Chomsky (1965) that have shaped mainstream generative grammar ever since: (1) Syntactocentrism, i.e. the assumption that all generativity is located in the syntactic component. Jackendoff argues that phonology and semantics have autonomous structures, only partially homomorphous to syntax. He proposes as an alternative a ''Parallel Architecture'' of the language faculty, where phonology, syntax and semantics have their own sets of formation rules and the resulting structures are linked by interface components. (2) The lexicon/grammar distinction taken from traditional grammar and formal logic, i.e. the assumption that there is a strict distinction between words (the locus of irregularities) and rules (regularities). In Jackendoff's Parallel Architecture a word is a type of interface rule that establishes a partial correspondence among pieces of phonological, syntactic and semantic structure. This conception is shown to be especially useful for the analysis of idioms. Finally four ways are pointed out in which the Parallel Architecture helps to integrate linguistics with cognitive neuroscience: First, the kind of organization proposed for the language faculty (i.e. several semi-independent combinatorial systems with linking interface principles) is plausible for other areas of the mind as well, whereas the syntactocentric architecture shows no resemblance to the rest of the mind. Second, the parallel architecture, being inherently non-directional, permits a closer relation between competence and performance theories. Third, the parallel architecture can integrate Conceptual Semantics, which applies the mental stance to meaning (in contrast to ''realist'' semantics in the tradition of Frege). Fourth, in contrast to the syntactocentric framework, the parallel architecture suggests a plausible scenario for the incremental evolution of language: Semantics was the first generative component of language to emerge, then meanings were paired with vocalizations (''paleolexicon''), and finally rules for concatenating words developed. In sum, the parallel architecture tries to accomplish what early generative grammar promised but failed to deliver: to integrate linguistic theory internally and more comprehensively with the mind/brain.
Chapter 3, ''Conscious and Unconscious Aspects of Language Structure'', is an exploration of the relation between language and consciousness. Two central hypotheses are that language is experienced as perceived sound - ''When one is experiencing language, the forms in awareness [...] most closely mirror phonological structure'' (p.81). - and that ''We are aware of the content of our linguistically expressed thoughts only by virtue of experiencing phonological images associated with them, plus other images that are inferentially nonefficacious'' (p.83). These hypotheses challenge certain fashionable views of consciousness, e.g. they entail that consciousness is not an executive capacity and does not involve higher-order thought. It is suggested that consciousness is closely connected to attention, which is anchored to mental structures in the perceptual layer. In addition to phonological structure, there are conscious ''valuation features'' that add a felt character to the entities in experience. Some candidates are [+/- external] (e.g. percepts are external, images are not), [+/- self-initiated] (e.g. hearing one's own voice vs. hearing someone else's voice), [+/- familiar] and [+/- affective].
In Chapter 4, ''Shaking Hands and Making Coffee: The Structure of Complex Actions'', the tools of theoretical linguistics for analyzing the language faculty are used to sketch an analysis of the human capacity for complex action. The analysis is carried out in some detail for two ''banal complex actions'' (p.111), namely shaking hands and making coffee. These complex actions consist of sequences of hierarchically ordered subactions (visualized in tree format, with Preparation, Head and Coda parts). Shaking hands involves two cognitive domains: the physical (moving the hand in space) and the social (knowing when it is appropriate to shake hands). Various parallels to language are pointed out: There is an enormous lexicon of basic actions (sometimes called ''Actionary''), structured in terms of inheritance hierarchies, and principles for composition (''grammar'') that involve instantiation and binding of variables and give rise to complex embedded constituent structure.
Chapter 5, ''Cognition of Society and Culture'', explores social cognition, our ability to understand and engage in social interactions. Social cognition is argued to involve - like language - a largely unconscious combinatorial rule system that is acquired with only imperfect evidence, thus based on inner unlearned (maybe even domain-specific) resources. Within linguistics, social cognition is of crucial importance for semantic analyses of predicates like ''to request'', ''to buy'', or ''to own'' which rest on basic social concepts like dominance, authority, obligation, value or agreement. Central to social cognition are relations to other persons, e.g. affiliations like kinship, alliances, and especially group membership. One of the ''axioms of groups'' is that members of the same group tend to behave favorably towards each other compared to non-members. A group is often conceptualized as a ''(super-)individual'' whose identity conditions can be independent of its members. Cooperation and competition among persons involve joint actions and joint intentions, which might be specific to humans, but have evolutionary precursors like joint attention and sexual behavior. Another important aspect of social cognition are normative rules of various kinds (moral rules, legal codes, religious codes). Jackendoff argues against the proposal that cognitive science should look for universal principles of moral judgment, partly because the distinction between social conventions and what is considered ''genuine morality'' is unstable across cultures. He suggests (by analogy to Universal Grammar) that we should rather study systems of norms as a whole and ask what innate resources children must have in order to learn them.
Part II, ''The Structure of Social Cognition and Theory of Mind'', consists of more detailed explorations of certain topics mentioned in the fifth chapter. The overall goal is to develop formal analyses of concepts involved in social cognition within the framework of Conceptual Semantics (cf. e.g. Jackendoff (1983)) which is briefly summarized at the beginning of the sixth chapter. The basic idea is that there is a level of mental structure, ''conceptual structure'', independent from syntax and phonology and epistemologically prior to language, that encodes the meanings of linguistic expressions. Conceptual structures are built combinatorially by a formal generative system and are linked to both linguistic structures and mental structures involved in perception and action by means of interface rules. Conceptual structure consists of various tiers (e.g. propositional tier, information structure tier, referential tier). The tier most relevant for the following analyses is the ''macrorole tier'' (formerly called ''action tier'') that contains conditions like ''X AFF Y'' (''X affects Y'') or ''X EXP Y'' (''X experiences Y'').
Chapter 6, ''Perception Verbs and Theory of Mind'', addresses the difference between verbs like ''see'' that treat perception in terms of the experience of the perceiver, hence involve theory of mind, and verbs like ''look'' that treat perception in terms of observable exploration of the environment, hence do not involve theory of mind. This difference is captured at the macrorole tier: The semantics of ''look'' involves the condition ''X AFF'' in the macrorole tier, whereas ''see'' involves ''X EXP Y''. Unlike ''AFF'', the function ''EXP'' constitutes a part of theory of mind in that it allows one to attribute experiences to others. This analysis suggests that theory of mind is not a ''module'' in the mind, but ''arises simply from having additional predicates such as EXP in the level of conceptual structure'' (p.207). The macrorole tier plays an important role in linking conceptual structure to syntax. However, the co-existence of Experiencer-subject sentences (like ''Sam regards this as wonderful'') and Stimulus-subject sentences (like ''This strikes Sam as wonderful'') poses a well-known problem for linking syntax and semantics (especially thematic roles). Jackendoff argues against the mainstream assumption of universal linking hierarchies and assumes that ''each verb that has EXP in its meaning must individually specify which macrorole is mapped to subject'' (p.211).
In Chapter 7, several types of ''Objective and Subjective Psychological and Evaluative Predicates'' (in English) are distinguished according to the following parameters: (a) syntactic category (verb vs. adjective), (b) Experiencer-subject vs. Stimulus-subject, (c) (overt) expression of the second argument. While the formal treatment of Experiencer-subject predicates is rather straightforward, two analyses of Stimulus-subject predicates are considered: The first one reduces them to Experiencer-subject predicates with the help of the formal device of lambda abstraction (e.g. the analysis of ''Golf is interesting to Bob'' can then be paraphrased as ''Golf is such that Bob is interested in it''). This analysis is ultimately rejected in favor of a formally simpler one, where Stimulus-subject predicates (e.g. ''Golf is interesting to Bob'') and Experiencer-subject predicates (e.g. ''Bob is interested in Golf'') are represented by formally identical structures at the thematic tier (''BOB BE [INTERESTED (GOLF)]''), but differ at the macrorole tier where the argument of EXP is marked that ends up in subject position. Another important parameter is whether the second argument is overtly expressed: Experiencer-subject predicates differ in whether the Stimulus is obligatory at the level of conceptual structure (e.g. ''interested'', ''amazed'') or optional (e.g. ''bored'', ''excited''). If the Stimulus is present in conceptual structure, they express intentional feelings, otherwise they express inherent feelings, a distinction well-established in psychology. Similarly, the second argument of Stimulus-subject predicates may not be expressed, in which case an ''objective'' or ''perspective-free'' interpretation results (cf. ''The problem doesn't interest me'' vs. ''The problem isn't interesting''). This reading is formally captured by introducing a ''generic individual'' YA as the first argument at the thematic tier and omitting the first argument of EXP at the macrorole tier (i.e. the Experiencer).
Chapter 8, ''Intending and Volitional Action'', deals with folk theory of mind and the semantics of attitude predicates. A basic distinction is drawn between situational attitude verbs (e.g. ''believe'', ''imagine'') that are typically complemented by that-clauses and pose no semantic restrictions on their complement (any situation will do) and actional attitude verbs (e.g. ''intend'', ''be willing'') that are typically complemented by infinitival clauses and pose certain semantic restrictions on their complement: it must denote a self-initiated and nonpast-directed action (where actions are events that involve an actor). The similarities and differences between these two types of attitude verbs are captured in the formal analysis: Both involve a predicate ''COM'' at the level of conceptual structure that is a conceptualization of the valuation feature [+/- committed] (familiar from chapter 3), but they differ in the semantic type of the second argument of COM: [Situation,-Action] (i.e. events viewed as pure events) in case of situational attitudes, [Situation,+Action] (i.e. events viewed as actions) in case of actional attitudes. The remainder of the chapter focuses on intention and offers formal conceptual analyses of intentional action, Dennett's intentional stance and joint intentions.
Chapter 9, ''The Logic of Value'', investigates how humans conceptualize values, using mainly linguistic expressions of value as evidence. The overall hypothesis is that value is a ''conceptualized abstract property'' connected to objects, persons and actions that plays a role in various rules of inference. A system of values consists of three basic elements: input rules (for assigning values to entities), output rules (for the selection of actions on the basis of values) and internal rules. Eight different, but partly related, types of value are distinguished: affective value, utility value, resource value, quality, prowess, normative value, personal normative value, and esteem. The same event can have positive value of some kind and negative value of some other kind, e.g. eating broccoli may be good for Bill (positive utility value) but he may hate it (negative affective value). Many values have subjective and objective versions (e.g. ''Eating broccoli is good for Bill'' vs. ''Eating broccoli is good''). The analysis of the latter is parallel to the treatment of psychological predicates without overt Experiencer in chapter 7, involving the ''generic individual'' YA.
The goal of Chapter 10 is to formalize the basic social notions of ''Fairness, Reciprocity, and Exchange'' using the analysis of value developed in the previous chapter. First, a formalization of fairness and of Fiske's (1991) four basic structures within the human social cognitive capacity for distributing goods, labor and responsibility (Communal Sharing, Authority Ranking, Equality Matching and Marked Pricing) is given in terms of conceptual structures. Second, Jackendoff turns to the concept of reciprocation and its expression in English by ''for'' constructions, as in ''Fred cooked Lois dinner for fixing his computer''. It is shown how a couple of invalid but ''seductive'' reasoning steps lead from a normative principle of reciprocation (''If X does something good for Y, then it is good of Y to reciprocate'') to a principle of deserved reward (''Good people deserve to be rewarded'') that can act as a grounding for religion. Third, the concept of (fair) exchange is analyzed, involving the joint task of agreement, theory of mind and cheater detection. While exchange as a concept is symmetrical, its expression in language is typically asymmetrical. For example, the English words ''sell'', ''buy'' and ''pay'' involve a common conceptual frame of transaction, but foreground different parts.
Chapter 11, ''Rights and Obligations'', investigates how people conceptualize situations in which someone has a right or an obligation. Following recent work in deontic logic, rights and obligations are formally analyzed as relations between (at least) two arguments: an Actor and an Action (obligations additionally involve a person to which the Actor is obliged). Rights and obligations are understood as social states (not propositional attitudes) that come in various kinds (e.g. existentially vs. universally quantified). An essential property of rights and obligations is that noncompliance has social or contractual consequences: In the case of obligations, ''If at [some time] t1 [a person] X has an obligation to [a person] Z to perform [some action] ACT1, and X does not perform it by [some later time] t2, then Z has the right at t2 to do something of negative [affective/utility-]value to X in retaliation for nonperformance'' (p.346). Jackendoff also addresses the questions of authority (who can impose an obligation on whom?) and of the acquisition of the social concepts of rights and obligations.
Chapter 12, ''Trumpets and Drums'', locates the investigation of social cognition undertaken in the preceding chapters in the broader context of cognitive science. Jackendoff criticizes that theoretical linguistics and wet neuroscience (as well as symbolic and connectionist models of the mind in general) tend to compete and ignore each other's results. He calls for more cooperation and emphasizes his aim to facilitate productive dialog by a reorientation of theoretical linguistics. Finally, Jackendoff turns to the general issue of what aspects of social cognition and theory of mind are special to humans. He emphasizes that theory of mind and social cognition should not be thought of as ''all-or-nothing modules'' in the mind, but as collections of certain abstract concepts that invoke the social plane (e.g. EXP and COM, discussed in chapters 6 and 8). Some of these concepts are specific to humans, some are also present in other primates, but possibly less highly differentiated.
EVALUATION This book is an immense effort to offer an integrated perspective on the (symbolic) mind from a linguistically based perspective, but drawing on insights from various disciplines including philosophy (ethics, logic, philosophy of language), (developmental) psychology, artificial intelligence, robotics, (evolutionary) biology and cognitive anthropology. One explicit goal of the book is to demonstrate the broad applicability of the theory of Conceptual Structure to all sorts of cognitive phenomena, and Jackendoff indeed manages to present a coherent and immensely comprehensive theory of the mind and mental structure, building on his rich earlier work, most notably Jackendoff (2002). What is new in the present book is the application of his framework to hitherto neglected areas of cognition, especially social cognition.
While broadness of coverage inevitably implies less attention to details, Jackendoff is eager to do his best and provide formalizations wherever possible, especially in the second part of the book. Since he always provides non-formal paraphrases, the arguments and analyses also remain easily accessible to readers less interested in deciphering formulas. However, for readers who like to deal with formal and logical systems, the offered analyses may not be completely satisfying. One reason is that Jackendoff assigns no interpretation to his formulas (conceptual structures). For him, this is not a bug but a feature, since he ''reject[s] all talk of the 'intentionality of mental representations,' the idea that mental structures are 'about' the world in some direct sense'' (p.6). This position is not only philosophically controversial (to say the least, cf. e.g. Gross (2005) for discussion), but also raises more specific questions. For example, exactly what sort of concept is ''YA'' that is said to ''stand for [a] generic individual'' (p.214) and how does it relate to recent theories of genericity? Somewhat troubling is also the intuitive use of the formal device of lambda abstraction without adapting it for conceptual structures by means of an explicit definition. However, in most cases the author-intended absence of formal interpretations of conceptual structures does not impair the insightful analyses (and nothing prevents the interested reader from constructing his own model-theoretic interpretation, e.g. along the lines of Zwarts and Verkuyl (1994)).
Since ''language'' is only one of the three catchwords that constitute the title of the book, one may wonder in what way this book is of interest specifically to linguists. First of all, it does a wonderful job in locating the study of language within the general enterprise of investigating cognition and the mind/brain. Second, it shows that in order to determine what is special to language, one has to know quite a lot about other mental components, e.g. the human capacity for complex action that is examined in chapter 4. Third, Jackendoff offers detailed semantic analyses of various English expressions, including perception verbs, psychological predicates, propositional attitudes etc., and shows how they depend on theories of other mental components (e.g. social cognition).
In sum, this book contains a fascinating and impressively comprehensive exploration of various aspects of cognition. It can be seen as an extension of Jackendoff's (2002) investigation of the language faculty to other areas of the mind, with a focus on social cognition. Of course, one cannot expect definite answers to all of the issues that are raised, but the book offers a coherent, sometimes deliberately quite speculative, but always very clear perspective on mental functioning that provides a fertile ground for further interdisciplinary research.
REFERENCES Chomsky, Noam. (1965) _Aspects of the theory of syntax_. Cambridge, Mass.: MIT Press.
Fiske, Alan Page. (1991) _Structures of social life: The four elementary forms of human relations_. New York: Free Press.
Gross, Steven. (2005) The nature of semantics: On Jackendoff's arguments. _The Linguistic Review_ 22: 249--270.
Jackendoff, Ray. (1983) _Semantics and cognition_. Cambridge, Mass.: MIT Press.
Jackendoff, Ray. (2002) _Foundations of language: Brain, meaning, grammar, evolution_. Oxford: Oxford University Press.
Zwarts, Joost and Verkuyl, Henk. (1994) An algebra of conceptual structure: An investigation into Jackendoff's conceptual semantics. _Linguistics and Philosophy_ 17(1): 1--28.
ABOUT THE REVIEWER Mathias Schenner is a PhD student in the EU project CHLaSC (Characterizing Human Language by Structural Complexity), coordinated by Manfred Krifka and Uli Sauerland, at the Center for General Linguistics, Typology and Universals Research (ZAS) in Berlin. His research focuses on evidentials, epistemic modals and perspective shifts in embedded contexts. More broadly, he is interested in formal models of language understanding.