The two target articles defend mental models against the minimalist criticism that they are theoretically unnecessary empirically unsupported. Garnham and Glenberg & Mathew (G&M) make a strong case for mental models by specifying the conditions for constructing them and by challenging misconceptions attributed by McKoon & Ratcliff to mental models. In addition, G&M's study broadens the scope of previous results and reinforces their original mental-models theory. The critics of mental models must re-examine the salience notion, develop a formalism to handle the range of effects taken as support for mental models, and recast the minimalist-constructivist debate.
2. Empirical support for the mental models notion has come from studies on readers' processing of the spatial attributes of brief texts. The results of Glenberg, Meyer, and Lindem (1987) are prototypical: Readers tended to foreground target objects more when they were spatially associated with a central character than when they were dissociated. One could quibble with one or the other minor point, for example, that the passages were relatively simple and homogeneous. However, no persuasive alternative account for these data has been offered. It is now necessary to advance our understanding of inferences and the mental models framework by addressing several issues, including those summarized in this commentary.
3. Mental model theorists have a clear preference for the spatial domain (e.g., Garnham, 6.1; Glenberg & Mathew, 1.1.2; but see Glenberg & Langston, 1992). The spatial dimension is fundamental even when it is not immediately apparent, as Morrow's (1986) penetrating analysis of grammatical morphemes and discourse structure suggests. Nevertheless, spatial relations constitute only a subset of the events, objects, and processes described in texts. Consider causal relations in texts: The research evidence for the validity of causal inference processes is growing (e.g., Keenan, Baillet & Brown, 1984; van den Broek, 1990). Examples of causally connected texts include such transparent ones as (see Keenan et al., 1984):
Racing down the hill, Joey fell off his bike. The next day he was covered with bruises.
There are also cases that require specific domain knowledge, as the following sentence illustrates (see Garnham, 1992, 5.1):
Chlorine compounds are frequently used as propellants because they do not react with other substances.
Whether the causal relation is readily understood, as in the Joey episode, or not, as in the chlorine sentence, it is not clear how mental models would capture such causal links. One could invoke one of the companion representations allowed by the mental-models framework. Doing so, however, means losing the advantage of mental models to capture the situation the text is about.
4. Mental modellers need to specify the format of mental models and of the information added by reader inferences. There is agreement that mental models are neither proposition lists nor images. It is not clear whether they can be thought of as arrays, a kind of distributed representation, or as a combination of several representations. It would be helpful to have such formulations, models of the mental model, so to speak. Such models would advance theory development and experimentation on mental models, much as they have done in other disciplines. The advantage of the companion representations, especially propositional text bases, is that they are empirically tractable. This is the reason that McKoon & Ratcliff "still characterize an inference as an encoded proposition" (see Garnham, 6.1).
5. Questions on inferences in reading will continue to attract attention: What types of inferences are there? When are they made, if at all? and, What specifically do they add to the reader's representation(s) of the text? I worry that minimalists and nonminimalists may both be right, and that we may have overlooked a critical dimension of inference making, namely, the cost-benefit dimension. According to Vonk and Noordman (1990, p. 462), "reading is a process in which a balance between costs and benefits is achieved." The benefit is the information the reader gains from reading. Costs result from the mental effort required by inferences. When texts are easy and short and comprehension requirements are minimal, as in most of McKoon & Ratcliff's passages, readers minimize inferences. On the other hand, if the comprehension demands are higher, for example, when the reader is asked to explain a technical concept, readers are more likely to draw inferences (Vonk & Noordman, 1990, Experiments 3, 4). Adding the cost-benefit continuum complicates matters, much as it has in psychophysics and in research on speeded performance. For example, rather than asking what the constraints are on inferences in reading, investigators would ask what the constraints are for a given point on the cost-benefit continuum.
6. Part of the appeal of minimalism lies in the minimal theoretical overhead used to account for a corpus of data. Ever since the Kintsch & van Dijk model was first introduced 15 years ago, the propositional text base fulfilled this function. It has proven to be a fertile stimulant for research as well. However, the text-base account of inferences in reading did run into problems, for example, vis a vis Glenberg, Meyer, and Lindem's (1987) demonstration that the associated/dissociated variable predicted foregrounding better than the text-base theory (see 1.0 above). McKoon & Ratcliff interpret the Glenberg et al. results in terms of the salience of the target: An object is salient to the extent that it is relevant to the discourse topic. Glenberg & Mathew point out, correctly, however, that salience may well be a derivative of a mental model. They also report data indicating that the associated/dissociated variable overrides the salience factor (see Glenberg & Mathew, 2.8.4).
7. Garnham and Glenberg & Mathew make a strong case for mental models. The two papers strengthen the conceptual basis of the mental-models approach by specifying the conditions for constructing mental models and by challenging two misconceptions attributed by McKoon & Ratcliff to mental models: that they are formed automatically and that they are" full representations of the real-life situation in the text." In addition, Glenberg & Mathew's study broadens the scope of the Glenberg et al. (1987) results and reinforces their original mental-models theory. There are questions the mental modelers need to address, such as the two sketched here: the mental-models framework should be extended to attributes of texts other than the spatial dimension (see 2. above) and a notation should be found for the modeling of mental models (3.).The critics of mental models should (i) re-examine the salience notion for its usefulness to account for the Glenberg et al. (1987) results, (ii) develop a formalism to handle the range of empirical effects taken as support for mental models, and (iii) recast the minimalist-constructivist debate, perhaps in terms of a cost-benefit tradeoff for inferencing.
Garnham, A. (1992) Minimalism versus constructivism: A false dichotomy in theories of inference during reading. PSYCOLOQUY 3(63) reading-inference-1.1
Glenberg, A. M., and Langston, W. E. (1992) Comprehension of illustrated text: Pictures help to build mental models. Journal of Memory and Language, 31, 129-151
Glenberg, A. M., and Mathew, S. (1992) When minimalism is not enough: Mental models in reading comprehension. PSYCOLOQUY 3(64) reading-inference-2.1
Glenberg, A. M., Mayer, M. & Lindem, K. (1987) Mental models contribute to foregrounding during text comprehension. Journal of Memory and Language, 26, 69-83
Keenan, J., Baillet, S. & Brown, P. (1984) The effects of causal cohesion on comprehension and memory. Journal of Verbal Learning and Verbal Behavior, 23, 115-126
McKoon, G., and Ratcliff, R. (1992) Inference during reading. Psychological Review, 99, 440-466
Morrow, D. G. (1986) Grammatical morphemes and conceptual structure in discourse processing. Cognitive Science, 10, 423-455
Van den Broek, P. (1990) The causal inference maker: Towards a process model of inference generation in text comprehension. In D. A. Balota, G. B. Flores d'Arcais & K. Rayner (Eds). Comprehension processes in reading. Hillsdale, N. J.: Erlbaum
Vonk, W. & Noordman, L.,G.,M. (1990) On the control of inferences in text understanding. In D. A. Balota, G. B. Flores d'Arcais & K. Rayner (Eds). Comprehension processes in reading. Hillsdale, N. J.: Erlbaum