The cognitive sciences are not only concerned with discovering the facts about cognitive abilities like mental imagery, it is also interested in explaining how the abilities work. Explanations can take different forms, and across the textbook we will encounter explanations ranging from (a) relatively simple claims about how something might work to (b) well-developed verbal theories to (c) highly specified computer simulations that propose working algorithms for specific cognitive abilities. Throughout, I encourage you to think about what kind of account would be a satisfying explanation of how a cognitive ability works. 

I think I can safely say that, as of right now, there is no broadly accepted theory or explanation about how mental imagery works. As we already discussed, there isn’t great consensus on the features of mental imagery, so the absence of a theory isn’t surprising. However, there has been a great deal of theoretical debate about mental imagery, and this debate provides a really nice example to discuss how theories and explanations are used in cognitive research.

Mental Imagery as explanation

In the previous section, I mentioned that we skipped about a hundred years of research. And there were some general trends in psychology that shaped how researchers asked questions about mental imagery. Some of the general trends correspond to gradual transitions between “schools” of thought in psychology.

At the beginning of psychology in the USA, Titchener (1867-1927) developed the “Structuralist” school of thought, and used careful, but subjective, introspection techniques to interrogate and discover individual components of cognition.

In the 1940s, Edward Tolman preempted the contemporary cognitive psychology research cycle strategy, but employed rats as participants. Tolman proposed cognitive maps (spatial memories) and provided evidence that even rats can form these spatial memories (e.g., Tolman et al., 1946). Humans make them too; if you rearrange your furniture and are navigating your home in the dark, watch out. Your cognitive map may need updating!

Behaviorism was another school of thought that favored asking questions that could only be answered with objective measures of behavior. Some “hard-core” behaviorists like B.F. Skinner (1904-1990), claimed that internal processes of the mind were simply outside the boundaries of scientific inquiry, and that psychology should only study observable phenomena like behavior. While behaviorists dominated American Psychology in the early-mid 20^th century, whole topics like mental imagery received less attention by researchers. However, in the 1950s, 60s, and 70s, there was a “cognitive revolution” of sorts, and more psychologists returned to asking questions about cognitive abilities, including mental imagery.

When mental imagery returned as a research topic it also came back as a potential explanation of other cognitive abilities. For example, by the 1960s there was already a very large literature on human memory abilities, which was often focused on examining factors that influence how well people can remember verbal stimuli like words. In 1963, Allan Paivio (Paivio, 1963) considered the possibility that mental imagery might be involved in tasks where people attempt to remember words from a list. He suggested that experiencing strong mental imagery when reading a word could make it easier to remember that word later on. Conversely, other words might not be associated with strong mental imagery, and those kinds of words might be harder to remember later. Paivio also provided some experimental evidence that was consistent with the idea that mental imagery impacts memory abilities.

Paivio’s concrete versus abstract memory task

Paivio used a standard paired-associate learning task. He ran experiments on elementary school students and college students, finding similar results. All the participants were given pairs of words to remember for a later memory test. Each pair involved an adjective and a noun, like “Ingenious-Inventor”. Importantly, there were two different kinds of word-pairs, and this was the critical manipulation in the experiment. The manipulation was whether the noun was more concrete or more abstract. Examples of the two kinds of word pairs are presented below:

What makes a noun more concrete or abstract? The general idea is that concrete words are more evocative, meaningful, and easier to mentally image than abstract words. For example, hearing or reading the word “Magician” might cause you to think of a colorful magician’s hat, whereas the word “Discourse” might not bring to mind specific mental images. Paivio chose words that he considered more concrete and more abstract when constructing the lists for his experiment.

During the encoding phase, the experimenter read lists of 16 word pairs out loud, with a two second pause in between. Half of the word pairs had a concrete noun, and the other half had an abstract noun. During the memory test, the experimenter read out only the first word from each pair (the adjective), and participants were asked to remember the word it was paired with and write it down. If you heard the word Amazing and you were given the concrete pair, then the correct answer would be to write down Circus. If you were given the abstract pair, then the correct answer would be Crusade.

The question was whether people would have better memory for the concrete nouns compared to the abstract nouns, and this is exactly what Paivio found. In the second experiment, he found that people remembered on average about 4.5 words correctly if they were concrete nouns, but only about 2 words correctly if they were abstract nouns. This is only a difference of 2.5 words, but the result seemed to be consistent across 120 university students, and it suggested that something about the concrete versus abstract quality of these words caused differences in memory performance for the words.

Paivio’s explanations 

Paivio entertained different explanations of his results, and the way he related results to explanations is fairly common in cognitive psychology. In the introduction of his paper he referred to new ideas about how memory might work from Miller, Galanter, & Pribram (1960), who suggested that mental imagery could help people efficiently organize, store, and then later retrieve information in memory. He also raised the possibility that mental imagery was the reason why participants remembered concrete (more image-able) better than abstract nouns (less image-able). Let’s call this the mental imagery explanation.

However, Paivio actually concluded that the “concept of mediating imagery… may be unnecessary” to explain his results. Instead, he described another possibility that memory performance was being determined by pre-existing associations between the word pairs. For example, some words are more likely to follow other words, and people may have different learned associations between different words. For example, it’s possible that I have a stronger learned association between “Noisy-Trumpet” (which was a concrete pair) than “Noisy-Gossip” (which was an abstract pair). If I was in this experiment and was given the cue word “Noisy”, I might be better able to remember “Trumpet” not because I formed a mental of image of a “Trumpet”, but because “Noisy” was already more strongly associated with “Trumpet” in the first place. Let’s call this the pre-existing association explanation.

The research cycle and theory testing 

More interestingly, Paivio ended his paper with a proposal for another experiment that could put these two explanations to the test. This is an instructive example of how the research cycle is used in cognition to refine questions from one experiment to another. A key ingredient is having at least two tentative explanations of the experimental results that make different testable predictions. Paivio’s were the mental imagery explanation and the pre-existing association explanation. The next ingredient is researcher creativity to come up with a new experiment capable of testing such predictions.

For example, imagine that there was a drug that immediately turned off everyone’s mental imagery. If mental imagery is responsible for people remembering concrete nouns better than abstract nouns, and if we repeated the experiment but gave people the drug to turn off their mental imagery, then what should happen to memory for the words? According to the mental imagery explanation, the difference in memory performance between concrete and abstract words should disappear. This is because mental imagery would be “turned off,” and would not be able to cause any differences in memory between the words. If there were still differences in memory between the words, this would be good evidence against the mental imagery explanation. (Relatedly, one might expect people with aphantasia to show no advantage for concrete words.)

Paivio didn’t have a magic wand to make mental imagery turn off, so he suggested a different approach, to control for aspects of the pre-existing association explanation. He proposed a new experiment could be repeated with random pairings between adjectives and concrete versus abstract nouns. For example, Noisy-Trumpet isn’t very random because the adjective noisy happens to be meaningful for some trumpets, and there could easily be pre-existing associations between Noisy and Trumpet. However, the influence of pre-existing associations could potentially be eliminated by randomly assigning unrelated adjectives to the nouns. If the advantage for remembering concrete over abstract nouns persisted under these new control conditions, then Paivio would have evidence that pre-existing associations do not explain his findings. And, perhaps the concept of “mediating imagery” would be necessary to explain these influences on memory for words.