Alan Garnham (1993) Dichotomy or not Dichotomy?: That is the Question. Psycoloquy: 4(16) Reading Inference (10)

Volume: 4 (next, prev) Issue: 16 (next, prev) Article: 10 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 4(16): Dichotomy or not Dichotomy?: That is the Question

Reply to Keenan on Garnham on Reading-Inference

Alan Garnham
Laboratory of Experimental Psychology
University of Sussex
Brighton BN1 9QG, UK


Keenan's conclusion that minimalism and constructionism can be contrasted with one another, is based on a misunderstanding of what I meant by claiming that minimalism vs. constructionism is a false dichotomy in theories of reading inference.


constructionism, inference, mental models, minimalism, reading, text comprehension.


1.1 No one can quibble with Keenan's (1993: 2.2.1) claim that if two theories are similar in one or more respects they may still differ in others. However, her application of this idea to the case of minimalism vs. constructionism, and her conclusion that the two can, after all, be contrasted with one another, is based on a misunderstanding of what I (Garnham, 1992) meant by claiming that minimalism vs constructionism is a false dichotomy in theories of inference during reading.


2.1 My point was that the dichotomy is a false one because minimalism and constructionism aren't theories about the same thing. Minimalism is a theory specifically about inference making and, at heart, a theory about which inferences are made during reading. It has nothing to say about the processes by which inferences are made -- except that those inferences are made automatically, in a particular technical sense of that term. Furthermore, there is general agreement that this claim is at best unsubstantiated and more probably incorrect. Neither does minimalism have much to say about what constitutes an inference. For example, although it claims that inferences establishing local coherence are made, it makes no attempt to specify what counts as establishing local coherence.

2.1 Constructionism is, to the best of my knowledge, not a theory at all. The label "constructionist" (or, better, "constructivist") can be applied to certain theories that McKoon and Ratcliff (1992: M&R) contrast with minimalism, but the claim that constructive processes contribute to inference making is only part of those theories. More particularly, it is not a claim that explains which inferences are made but, rather, one that explains how certain inferences are made, or how they would have to be made, if they were made properly (see below). My own view (see e.g. Garnham & Oakhill, 1993) is that what is needed is not a theory of inference making per se, but a theory of text and discourse comprehension, namely, the mental-models theory. According to that theory, constructive processes, as defined by Bransford, Barclay and Franks (1972: see Garnham, 1992: 3.2 for details), are a crucial part of text comprehension. In particular, constructive processes contribute to the making of many inferences in the course of comprehension. Note, however, that the claim is that constructive processes contribute to inference making, not that they constitute it.

2.3 Having identified the constructive nature of processes that contribute to inference making, the mental models theory holds that when one considers many of the inferences that M&R agree are drawn "automatically" during reading, constructive processes contribute to the making of them. This is not so much an empirical claim as an analytic one -- based on a detailed examination of what the inferences comprise. For example, the local coherence of a text in which "John stirred his coffee" is followed by "The spoon was dirty and tainted it" is, for most readers, underwritten by the general knowledge that spoons are used to stir coffee. This combination of information explicit in the text with relevant background knowledge, to produce a representation of the situation that the text is most probably about, is the hallmark of constructive processing. Of course, analysis cannot show that readers engage in constructive processing when they read a text containing these two sentences. It cannot even show that local coherence between the two might not, in some sense, be established by nonconstructive processes (see below). Even when it has analysed inferences, a theory of text comprehension still has to make empirical claims about whether, or under what circumstances, such inferences are made. However, by recognizing the need to provide an account of what constitutes an inference, the mental models theory is well placed to consider the kinds of process that must underlie the making of particular inferences, if they are made.


3.1 It is useful to consider Keenan's four proposed points of contrast between minimalism and constructionism in more detail, since they help to highlight possible disagreements about what inferences are made. Her first contrast is between minimalism's claim that few inferences are made and constructionism's claim that many are. In fact, only minimalism makes any claim about how many inferences are made. Although adherents of constructionist theories have often suggested that many inferences are made as a matter of course during reading, such theories need make no such claim. Indeed, it is reasonable to extend M&R's suggestion about the strategic nature of inference making to all inferences. If M&R's automatic inferences are not automatic in the technical sense, there may be circumstances in which even those inferences are not made -- perhaps, for example, when readers gets to the end of a page of text and realize they have taken nothing in from it. My own hunch, admitted unsubstantiated, is that there is a mode of reading that corresponds to one described by what I (Garnham, 1992: 2.1) identified as the precursor to M&R's minimalism. That is to say, a mode in which readers make just those inferences that are necessary for a coherent interpretation of the text and no more.

3.2 Keenan's second contrast is between the local inferences of minimalism and the local plus global inferences of constructionism. However, minimalism does not claim that global inferences are unnecessary for a proper interpretation of a text. It does not even claim that they are never made as a matter of course, since apparent local incoherence can sometimes only be resolved by global considerations, and M&R imply that it will be so resolved. Thus, the question of local vs local and global inferences is just another one about which inferences are made when. Hence, it is a question about which constructionism per se need have nothing to say.

3.3 The third contrast is between theories that focus on a representation of the linguistic form of the text being read and those that focus on a representation of the situation described by the text. Keenan (1993: 2.1) agrees with my remark (Garnham, 1992: 2.1) that minimalism is not a new position but points out that it provides a useful reminder of a striking aspect of the data on inference in reading. In a similar vein, I would argue that mental models theory provides a useful reminder of what should be an obvious point about text comprehension. When we read a simple descriptive text about the real world or an imaginary one, the information we extract from the text is information about that world. If we read (and understand) a newspaper article about, say, the events in Sarajevo on a certain day, we set up a mental representation of those events -- a mental model of a situation in the world. We are not primarily interested (unless, for example, we are literary theorists) in the linguistic form of the text. And although detailed processing guided by the form of the text is needed to create a mental model, no viable theory of text comprehension can be based on the idea that the aim of the processes it describes is to create a representation of the linguistic form of the text. That this point is not always obvious to psycholinguists is partly explained by the history of their discipline, and its preoccupations.

3.4 Finally, Keenan claims that minimalism posits little use of background knowledge in inference making while constructionism holds that there is extensive use of such knowledge. This claim is difficult to interpret because the background knowledge that must be used to make an inference is determined largely by the pieces of information to be linked. The inference connecting "John dropped the delicate glass pitcher on the hard floor" and "he went to fetch a broom" inevitably makes use of background knowledge about what happens when delicate glass objects hit hard floors. One interpretation of Keenan's claim would be that minimalists believe that background knowledge is little used, since few inferences are made. But on this interpretation the fourth contrast is just a special case of the first. Another possibility is that minimalism claims that readers try to avoid invoking background knowledge by using alternative heuristic strategies to establish local coherence. For example, instead of using constructive processes to interpret pronouns, they might use strategies such as subject assignment (take the pronoun to refer to the subject of the previous clause) or parallel function (take the pronoun to have the same grammatical role as its clause and its antecedent in its clause). Such nonconstructive processes may indeed be used by readers, but they will inevitably lead to errors of interpretation in one of the sentences in pairs such as:

        Max confessed to Bill because he wanted a reduced sentence.
        Max confessed to Bill because he offered a reduced sentence.

3.5 Greene, McKoon and Ratcliff (1992) have indeed put forward a hypothesis of this kind about pronoun assignment, based on the idea that there is often a preferred antecedent for pronouns in the reader's "focus of attention." Furthermore, Greene et al. claim that pronouns are not properly interpreted when the focus of attention fails to determine a unique referent -- at least, they are not properly interpreted when readers rely on "automatic" comprehension processes. Ironically, however, the experimental findings Greene et al. use to support this notion may have a strategic explanation. Their subjects were not forced to read for comprehension and could readily perform the probe task by simply registering the main content words. They did not need to look at the pronouns at all.


Bransford, J.D., Barclay, J.R. & Franks, J.J. (1972). Sentence memory: A constructive versus interpretive approach. Cognitive Psychology, 3, 193-209.

Garnham, A. (1992). Minimalism versus constructionism: A false dichotomy in theories of inference in reading. PSYCOLOQUY 3(63) reading-inference-1.1

Garnham, A. & Oakhill, J.V. (1993). Modeles mentaux et comprehension du langage. In M-F. Ehrlich, H. Tardieu & M. Cavazza (Eds.), Les modeles mentaux: Approche cognitive des representations. (pp.23-46). Paris: Masson.

Greene, S.B., McKoon, G. & Ratcliff, R. (1992). Pronoun resolution and discourse models. Journal of Experimental Psychology: Learning, Memory and Cognition, 18, 266-283.

Keenan, J.M. (1993). Thoughts about the minimalist hypothesis. PSYCOLOQUY 4(2) reading-inference.3

McKoon, G. & Ratcliff, R. (1992). Inference during reading. Psychological Review, 99, 440-466.

AUTHOR NOTE: My work on mental models has been supported by ESRC grant C 0023 2439 "Mental models and the interpretation of anaphora". Thanks to Jane Oakhill for comments on an earlier draft.

Volume: 4 (next, prev) Issue: 16 (next, prev) Article: 10 (next prev first) Alternate versions: ASCII Summary