2.2 Reverse Engineering An Outline
Within a good paper you can find a good outline.
What structure should your outline have? You need an outline which helps you to write. You also need an outline which you can easily and usefully scaffold. To propose such an outline, I follow some common writing advice (Sarnecka, 2019; Strunk & White, 1959).
First, writing guides advise you to treat the paragraph as writing’s basic element because a good paragraph conveys a single topic (Strunk & White, 1959). Figure 2-1 illustrates ‘one paragraph, one topic’ by representing a paragraph with a cone. The cone’s single point reminds you the paragraph conveys only one topic.
Second, writing guides remind you about the importance of a paragraph’s first and last sentences (Sarnecka, 2019; Strunk & White, 1959). A paragraph’s first and last sentences convey a good paragraph’s topic. We call a paragraph’s first sentence the topic sentence. A topic sentence states the paragraph’s topic to your reader. We call a paragraph’s last sentence the concluding sentence. The concluding sentence brings the paragraph’s topic home by summarizing or restating the topic for your reader.
A paragraph’s topic sentence and concluding sentence relate meaningfully to one another because both convey the same topic. “One easy way to check for coherence in a paragraph is to read just the topic sentence and the concluding sentence. If they aren’t on the same theme, the paragraph has wandered off track and needs some attention” (Sarnecka, 2019, p. 236). Figure 2-2 uses arrows to illustrate the meaningful relationship between a paragraph’s topic and concluding sentences.
Writing guides offer a third kind of advice by encouraging you to find examples of other writer’s work to admire, to learn from, or to try and emulate. I now use the three pieces of writing advice to propose a desirable structure for an outline. I start with a writing sample I admire. I then use the other advice to ‘reverse engineer’ an outline which could create the writing sample. I do so by examining each paragraph’s topic and concluding sentences and by inferring a paragraph topic from the two sentences.
For my writing example, I chose ‘Connectionism and cognitive architecture: a critical analysis’, published in Cognition, and written by Jerry Fodor and Zenon Pylyshyn (Fodor & Pylyshyn, 1988). The paper criticizes cognitive scientists who use artificial neural networks. I chose the paper because it relates to my own neural network research, because of its influence (1792 citations as of February 2025), and because of its solid writing. You do not need any knowledge about cognition or connectionism to understand the reverse engineering I illustrate with the Fodor and Pylyshyn example.
How can I reverse engineer an outline from my writing sample? I pay heed to the paper’s paragraphs. I start by copying each paragraph’s first and last sentences. Table 2-1 provides the topic sentence and the concluding sentence for the first eight paragraphs of Fodor and Pylyshyn’s (1988) introduction.
After isolating the topic and concluding sentences, I next read each pair to figure out each paragraph’s topic. Remember, you state a paragraph’s topic with its topic sentence, and the paragraph’s concluding sentence should also refer to the paragraph’s topic. The final column in Table 2-1 provides my proposed topics for the introductory paragraphs in Fodor and Pylyshyn (1988).
Table 2-1 reverse engineers a paper which is important to my own writing and research experience. However, your interests likely differ from mine, making Fodor and Pylyshyn (1988) a poor example for you. Remember, though, you can perform the procedure I used to create Table 2-1 on a different paper, one related to your own interests, to produce an example more meaningful to you.
Table 2-1 illustrates a one-to-one mapping between each paragraph topic and each paragraph’s topic and concluding sentences. In the current chapter I propose a scaffold for developing an outline which includes paragraph topics; the scaffold helps me convert each paragraph topic into the paragraph’s first and last sentences. I begin describing my scaffold by considering the need to meaningfully organize paragraph topics.
| Table 2-1. The first and last sentence of the paragraphs which make up the introduction to Fodor and Pylyshyn (1988), and the assumed topic of each paragraph. | |||
|---|---|---|---|
| Paragraph | Topic Sentence | Concluding Sentence | Topic |
| 1 | Connectionist or PDP models are catching on. | There are also, inevitably, descriptions of Connectionism as a Kuhnian “paradigm shift”. | Connectionism is interesting |
| 2 | The fan club includes the most unlikely collection of people. | Almost everyone who is discontent with contemporary cognitive psychology and current “information processing” models of the mind has rushed to embrace “the Connectionist alternative”. | Connectionism attracts scholars unhappy with the status quo |
| 3 | When taken as a way of modeling cognitive architecture, Connectionism really does represent an approach that is quite different from that of the Classical cognitive science that it seeks to replace. | The style of processing carried out in such models is thus strikingly unlike what goes on when conventional machines are computing some function. | Connectionism attracts researchers unhappy with the status quo because connectionism departs from standard approaches |
| 4 | Connectionist systems are networks consisting of very large numbers of simple but highly interconnected “units”. | The behavior of the network as a whole is a function of the initial state of activation of the units and of the weights on its connections, which serve as its only form of memory. | What are the properties of connectionism which make it different? |
| 5 | Numerous elaborations of this basic Connectionist architecture are possible. | The term ‘Connectionist model’ (like ‘Turing Machine’ or ‘Von Neumann machine’) is thus applied to a family of mechanisms that differ in details but share a galaxy of architectural commitments. We shall return to the characterization of these commitments below. | Many different versions of connectionist networks are possible |
| 6 | Connectionist networks have been analyzed extensively – in some cases using advanced mathematical techniques. | Of even greater interest is the fact that such networks can be made to learn; this is achieved by modifying the weights on the connections as a function of certain kinds of feedback. | Connectionist networks have been studied a lot, using mathematical techniques, because they learn |
| 7 | In short, the study of Connectionist machines has led to a number of striking and unanticipated findings’ it's surprising how much computing can be done with a uniform network of simple interconnected elements. | Surely this is a proposal that ought to be taken seriously: if it is warranted, it implies a major redirection of research. | Studies of connectionism reveal surprising results: we should take connectionism seriously |
| 8 | Unfortunately, however, discussions of the relative merits of the two architectures have thus far been marked by a variety of confusions and irrelevances. | The arguments that then appeared to militate decisively in favor of the Classical view appear to us to do so still. | Twist: But, when you properly compare connectionist networks to classical models, classical models still win! |
