Zajonc, R. B., Feeling and Thinking: Preferences Need No Inferences. American Psychologist, 1980, 35, 151-175.

This paper is about the idea that not all feelings (or preferences) are based on cognitive processes but often precede them. Affect doesn't require extensive cognitive processing to occur.

Contemporary psychology sees feelings as after cognition. Before one can like something they must be aware of it and have assessed some critical attributes.

Yet feelings are a primary variable in interactions of people. The social interpretation of affect is as much non-verbal than verbal. People can differentiate the tone of voice much more reliably than the content (or even the language used!).

According to prevelant models for affect, preferences are formed after cognitive activity. Yets its difficult to separate thoughts (based on information) and feelings (based on energy transformations). The author believes that affect precedes cognition. We can be afraid of something before we are aware of its presence (though cognition will quickly change feelings from their initial state).

Other researchers have similar views. In decisions we often decide without extensive cognitive processes. Quite often "I decided in favor of X" is no more than "I liked X". The cognitive processes are used more for later justification than factor in the initial decision.

Affect is the first link is the evolution of complex adaptive functions that separated animals from plants. Unlike cognitive judgements, affective feelings often cannot be avoided. One can control the expression of emotion but not the emotion itself. Affect seems to be more easily noticed and recalled than thoughts. They are less controllable than thoughts.

Also, once feelings are created they are less likely to be changed than cognitions. Even aftercomplete removal of causal effects (as in subject debriefings), people feelings and opinions often remain. We rarely change our initial impressions of something, because we trust our reactions.

Feelings are more a reflection of oneself than of others. We also have trouble articulating our impressions (e.g., reasons for liking or disliking a person). While feelings rely mostly on non-verbal channels, they are very efficient. It seems that feelings are independent of any particular sensory modality.

Also, we have had extremely difficulting in predicting individual preferences from cognitive attributes, and we have also had limited ability to make attitudinal changes via communication and persuasion.

Therefore, if cognition isn't used to create affect, there must be some other factors that combine with affect and cause us to evaluate and experience attraction, repulsion, pleasure, and conflict quite early in sensory input. Zajonc defines these factors as preferenda.

Preferendas are unspecific -- some kind of interaction between gross object features and internal states of the individual.

Affective reactions are often separated from content. We can remember the feeling we had leaving a movie long after we forget what the movie was all about.

There is some evidence that preferences can come before inferences. Studies that show increasing preference with repeated exposure to an object are still unexplained by traditional theories. Also, studies with Japanese characters showed people had higher preference for symbols they had seen before, even when they didn't recall seeing them! Modeling of the data from this experiment suggests there is an affective path independent of recognition.

Another study where people heard random tones as they were listening and reading a book on tape preferred tone sequences that had been repeated before, even though they couldn't recognize the tone sequences overtly. Another study where people were exposed to polygon shapes for 1 millisecond were preferred over those with no pre-showing, thougn again people claimed not to have seen them the first time. The results seem to indicate there are some features (preferenda) that allow people to experience affect without formally gaining cognition.

Of course, some form of recognition must have occured, but it must be primitive or minimal, though not at a conscious level.

One study briefly flashed words followed by a mask. They then asked if they saw words before the mask, and if so they were showed word pairs and asked which word was 1. more visually similar 2. more semantically similar,

Surprisingly, with reduced stimulus time the first ability to disappear was to determine the presence of the words. The second was the physical similarity. The last was the semantic capabilities.

Feeling and Thought: Two Systems?
In another study, people were asked to listen for differnt things in word lists, and then later asked to recall the words. Those that were asked to rate each word for pleasantness were more likely to recall the words (vs counting letters or other schema).

Many other studies showed similar effects that using affective means to expose words or pictures allowed people to recall them better later. Also techniques that force self-reference (means the same as ___ ) of adjectives showed that not only did it improve recognition, but the speed of response was much faster.

One more studies showed better retention if people formed unpleasant or pleasant associations between word pairs than with neutral word pairs.

In all of these studies, the researchers usually postulate two separate cognitive systems, rather than consider an affective system.

It seems that pictures are recognized faster than words, maybe because they elict affect faster than abstractions like words do.

There also appear to be some physiological reasons that suggest separate affect /cognitive processing. Babies learn to show emotions long before language. Emotional responses were also present in pre-humans long before the evolution of language.