视觉语法的理论框架：从经验到科学

1.1 问题的提出：摄影是否拥有可形式化的语法？

Photography as a language — this metaphor has circulated for over a century, yet its precise analytical value remains contested. 当我们说一张照片"构图良好"（well-composed）、"节奏感强"（rhythmic）、"张力十足"（tensely composed）时，我们究竟是在描述一种可被验证的视觉结构，还是仅仅在投射一种文化习得的审美判断？The core question of this book — and the question with which this opening chapter must grapple — is whether photographic vision possesses an objective, formalizable, and teachable *visual grammar* (视觉语法) on par with the syntactic structures that govern verbal language.

The stakes of this question are both theoretical and practical. 在理论上 (theoretically), the answer determines whether image analysis can migrate from the humanities — where interpretation remains irreducibly subjective — into a domain capable of producing falsifiable propositions. 在实践上 (practically), if such a grammar exists, it implies that compositional mastery is not merely an incommunicable gift but a structured competence that can be decomposed, transmitted, and tested. Every photography pedagogy from the New Bauhaus school to the contemporary visual analytics labs of computational vision science implicitly assumes an affirmative answer. Yet few have articulated precisely what that grammar consists of, how it differs from a mere catalogue of rules, and where its limits lie.

This chapter lays the theoretical foundation for the entire book. We pursue three intertwined objectives. 第一 (first), we reconstruct the historical genealogy of visual proportion theory, tracing the trajectory from Leon Battista Alberti's geometric window of the Renaissance to Norman Brosterman's contemporary rediscovery of the Fibonacci structure in modernist photography. 第二 (second), we extract from Gestalt psychology (格式塔心理学) the experimental and theoretical apparatus that transformed visual organization from a metaphysical doctrine into an empirically testable phenomenon. 第三 (third), we propose a tripartite model of visual grammar — visual lexicon (视觉词法), visual syntax (视觉句法), visual semantics (视觉语义) — derived as an analogy from Ferdinand de Saussure's structural linguistics, and we argue that this model is not a metaphor but an analytical instrument of genuine explanatory power.

Before we proceed, a necessary terminological clarification. Throughout this book, *grammar* is used in its technical linguistic sense: a finite set of generative rules that produces well-formed structures from a finite repertoire of elements. A grammar is more than a list of conventions; it is a combinatorial system with productive capacity. Whether photographic images admit such a system — and, if so, at what grain of description — is the very question this chapter refuses to presuppose. We will treat it as an open problem and argue, by the chapter's end, that a qualified affirmative answer is defensible.

1.2 历史演进：从 Alberti 的透视窗到 Brosterman 的比例场

The history of visual proportion theory (视觉比例学说) is in many respects the prehistory of visual grammar. Although the term "grammar" is rarely invoked before the twentieth century, the enterprise of identifying recurrent numerical and geometric relationships in strong visual work — and treating those relationships as causally implicated in the work's perceptual power — is over five centuries old.

1.2.1 Alberti 与文艺复兴的"窗口"模型

Leon Battista Alberti's *De pictura* (1435), the founding treatise of linear perspective, established what we might call the *geometric-window model* (几何窗口模型) of pictorial construction. Alberti's instruction to the painter — to consider the picture plane as an open window through which the eye surveys a mathematically continuous world — implies three commitments that will haunt every subsequent theory of visual grammar. First, the picture is fundamentally planar (二维平面性): it is a finite surface on which a scene is projected. Second, the projection obeys determinate geometric rules: orthogonals converge at a vanishing point, distances are measurable through diminution. Third, the rules are normative: a painting that violates them is not merely unconventional but formally deficient.

Alberti's grammar, therefore, is essentially syntactic. He is not concerned with what the picture represents (its semantic content) or with the emotional effect on the viewer (its rhetorical reception), but with the rule-governed relationships among the elements of representation. The famous *velo* (veil) and the *costruzione legittima* (legitimate construction) are syntactic devices in precisely the structural-linguistic sense: they specify how primitives — lines, points, gradients — combine into well-formed figures.

The grammar Alberti codifies is, however, a grammar of *spatial representation* rather than of *pictorial composition* per se. It tells the painter how to depict three-dimensional space on a two-dimensional surface, but it says comparatively little about how the depicted elements should be arranged within that surface. The latter problem — what we would today call composition (构图) — is addressed only obliquely, through Alberti's recommendations on the placement of the principal figure, the use of *istoria* (narrative), and the proportional relationships among human bodies. It will take three more centuries before a systematic compositional grammar begins to emerge.

1.2.2 Peacham、de Piles 与巴洛克的比例学说

The transition from spatial representation to compositional organization is most clearly marked in the seventeenth century, particularly in Henry Peacham's *The Gentleman's Recreation* (1622) and Roger de Piles' *Cours de peinture par principes* (1708). Peacham, writing in the English tradition of emblem books, draws an explicit analogy between the structure of the painted picture and the structure of the oration: there is to be an *inventio* (invention of subject matter), a *dispositio* (arrangement of parts), and an *elocutio* (execution in line and color). The *dispositio* in particular corresponds to what we now call visual syntax — the orderly arrangement of visual elements to produce intelligible and persuasive structures.

De Piles, a century later, formalizes this oratorical analogy into something closer to a numerical system. His *balances des peintres* assigns weighted scores to painters across four dimensions — composition, drawing, color, expression — and the resulting rankings (Raphael at the apex, Titian below Rubens, Caravaggio far below) instantiate, perhaps for the first time, a quantitative evaluation of compositional competence. Although de Piles' numerical method appears quaint to modern eyes, it represents a significant theoretical move: the assertion that visual structure can be measured, compared, and ranked along common dimensions. This is the move that will eventually make possible the transition from connoisseurship (鉴赏术) to visual analytics (视觉分析学).

1.2.3 Ruskin 与维多利亚时期的道德化视觉分析

John Ruskin's *The Elements of Drawing* (1857) and *Modern Painters* (1843-1860) introduce a different inflection. Where Alberti and de Piles were concerned with the geometry of representation, Ruskin is concerned with what we might call the *ethics of looking* (观看的伦理学). His grammar, insofar as he has one, is inseparable from a moral claim: that accurate perception of natural form is itself a moral act, and that compositional deviation from natural truth is a moral failure.

For our purposes, Ruskin is important not because his grammar is operationally precise (it is not), but because he establishes three propositions that subsequent visual analytics will have to either inherit or reject. 第一, Ruskin insists that visual rules are not arbitrary conventions but responses to features of the natural world — they have a quasi-empirical basis. 第二, he claims that violating these rules produces not just formal incoherence but perceptual and moral confusion — the rules have normative force. 第三, he treats the eye as trainable: visual grammar is learnable, not innate — it is a cultural acquisition with biolog