Developing conceptual semantics analyses
How are Conceptual Semantics analyses like these developed? As with any empirical investigation, there are no hard and fast rules. One simply starts with whatever presents itself – interesting questions, hunches, or existing analyses which seem to provide a promising starting point for further investigation. One such existing analysis is provided by the data in (5)–(8). In the spirit of Gruber (1965), Jackendoff points to the fact that there are apparent parallelisms in the use of the verbs go, be and keep and the prepositions to and from across four different semantic domains: spatial location and motion, possession, property-ascription and activity-scheduling (examples adapted from Jackendoff 1990: 25–26):
These sentences show the same verbs and prepositions operating in intuitively similar ways across the four semantic domains. Concentrating sim ply on the verbs, the sentences with go (the (a) sentences in (5)–(8)) express a change of some sort, with the end points of the change being expressed by sentences using the verb be (the (b) sentences).
In order to capture these intuitive similarities, Jackendoff claims that go and be each realize an identical conceptual meaning across all the sentences in which they appear. He expresses these identical conceptual meanings with the following formalism:
The important claim that Jackendoff makes about this analysis is that it also applies to all the other (a) and (b) sentences. The scheduling examples, for instance, are represented as in (11):
The other examples can be given similar analyses. All that differentiates them is a semantic field feature that specifies whether the concepts are applying to possession, property-ascription, motion, or whatever. (The change of semantic field also introduces variations in the expression of the Place argument: in (5b) and (8b) it is expressed by a prepositional phrase headed by in and on respectively; in (6b) by a possessive noun phrase (Philip’s) and in (7b) by an adjective phrase. These variations have to be explained by other mechanisms, which do not affect the point relevant here.)
What is the advantage of this representation? Here is Jackendoff’s own explanation:
The point is that at this grain of analysis the four semantic fi elds have a parallel conceptual structure. They differ only in what counts as an entity being in a Place. In the spatial fi eld, a Thing is located spatially; in possessional, a Thing belongs to someone; in ascriptional, a Thing has a property; in scheduling, an Event is located in a time period.
This notation captures the lexical parallelisms in [(5)–(6)] neatly. The different uses of the words go, . . . be, . . . , from, and to in (6) are distinguished only by the semantic fi eld feature, despite the radically different sorts of real-world events and states they pick out. (1990: 26)
BE and GO are core functions in the conceptual organization of events. Two other important ones are INCH and CAUSE. To illustrate them, let’s consider the three sentences in (12).
Sentence (12a) is the conceptually simplest, consisting simply of the state function BE with Thing and Property arguments, as diagrammed in (13):
To obtain (12b), we add the INCH function. This stands for ‘Inchoative’ (Latin: ‘beginning’), and denotes the coming into being of an event. The door opened thus receives the following analysis:
To get (12c), we can add the CAUS function. This stands for ‘causative’. John opened the door is analysed as simply involving the addition of this function to the previous structure:
As we have seen, this is one of the attractions Jackendoff claims for Conceptual Semantics analysis: the fact that primitives postulated for one area of meaning turn out to have a wide explanatory potential (1991: 42). We will end this short description of the framework by considering two other examples of this.
The first concerns the parallelism between the plural of nouns (mice, committees, funfairs), and iterative uses of verbs like those in (16):
Here flashed can only be understood as indicating a repeated series of individual flashes. It has often been observed that this ‘iterative’ meaning can be compared to the meaning of the plural in nouns: in both cases there is a multiplicity of entities – several mice, committees, funfairs; and several separate flashes. Jackendoff therefore proposes that plural and repetitive meanings correspond to the same element in conceptual structure, which he represents as PL. When added to a Thing element, PL creates a multiplicity of things, when added to an Event element, it creates a multiplicity of events (1991: 16).
A more complex example (1991: 31) also concerns a conceptual correspondence between spatial and temporal domains. Jackendoff proposes a 4-place dimensionality feature DIM which represents the conceptual underpinnings of dimensionality. Points are conceptualized as DIM 0D, lines/curves as DIM 1D, surfaces as DIM 2D, and volumes as DIM 3D. Consider line-like entities like roads, rivers, or ribbons. Jackendoff analyses our conceptualization of their spatial qualities as involving two DIM specifications. The most basic specification is DIM 1D: this expresses the conceptualization we have of such entities as single lines. (This conceptualization is reflected in the way rivers are represented on maps, for example.) But there is also a secondary dimension, the lateral cross-section dimension. This is the dimension which we attend to when we cross a road or river, or cut a ribbon. The two dimensions are shown in Figure 8.1.
The primary dimension may be conceptualized as bounded, as in a particular length of ribbon, or unbounded, as in a road that we think of as continuing indefinitely. The secondary dimension, by contrast, is always bounded: a ribbon has an edge, a road has a side, and rivers have banks. Boundedness is coded by the feature [±b]. There is also a feature [±i], which refers to ‘internal structure’ – whether or not the entity con sists of a ‘multiplicity of distinguishable individuals’ (Jackendoff 1991: 19). Aggregates (the entities usually expressed by plural nouns) will be [+i], singular nouns and nouns for substances (entities normally expressed by mass nouns), will be [–i] (1991: 19–20). This gives us the following specification for the conceptual structure underlying the nouns road, river and ribbon in the singular. The inner brackets refer to the secondary dimension:
Adding the [PL] operator has the function of changing the [i] value in the outer bracket to positive.
The dimensionality feature is not limited to space: it can be extended to time. Points in time (midnight, the moment I realized my mistake) and point events (the light turned on, they blinked, the climber reached the summit) are [DIM 0D]. Periods of time and states and events with duration are [DIM 1D]. Space and time, on this picture, are thus represented by identical conceptual primitives.
Any one-dimensional entity can also have a direction, which Jackendoff represents with the feature [DIR]. Ordinary lines, for example, which are directionless, lack this feature, but arrows and vectors possess it. The direction feature allows us to generalize between Places and Paths, which we introduced above as two of the major conceptual categories in Jackendoff’s system. Paths are conceptualizations like from the starting line to the finish or to the lighthouse: these are one-dimensional and directional, coded as [DIM 1D DIR]. Places, con trastingly, are non-directional, and can be regions of any dimensionality: at this point is zero-dimensional, along the line is one-dimensional, in the circle two dimensional and in the cup three-dimensional (Jackendoff 1991: 31).
Jackendoff then applies a similar analysis to events and states. Here are his own words:
I would like to extend this, a little speculatively, to the relation between Events and States. States like X is red or X is tall are conceptualized as ‘just sitting there’ – they have no inherent temporal structure. . . . Events, by contrast, do have an inherent temporal structure which proceeds in a definite direction. I would like to suggest therefore that the two categories be combined into a supercategory called Situation, with States as the undirected case and Events as the directed case. (1991: 31)
This representation shows the same characteristics as the other examples we have seen – it posits a small number of abstract conceptual constituents which underlie an apparently divergent range of different meanings.
