Social cognition is concerned with the mental representations and processes involved in social judgements and social behaviour. Bottom-up and top-down processing will be discussed in relation to the social cognitive processes; perception, encoding and storage, reference will be made to the role of schemas, heuristics and the effect of priming. Top-down explanations of processing rely on previous experiences and expectations, and highlight the importance of the context. In contrast, bottom-up explanations suggest social judgements and behaviours are directly influenced by sensory imput (Hogg & Vaughn, 1998).
The basic foundations of top-down and bottom-up processing have been adopted by many theorists and have been empirically demonstrated in the areas of; logical reasoning, person perception, person memory, judgement tendencies and linguistic communication. Research into the two processing modes has concentrated on three major components; how people process in quick and effortless fashion (top-down processing), how they process when willing and able to engage in extensive thought (bottom-up processing), and finally, what conditions encourage each type of processing.
Cognitive processes are staged. Observed stimulus events must be perceived, then encoded and stored. The encoding and interpretation of the perception is heavily influenced by prior knowledge stored in memory (Fielder & Bless, 2001). Newly encoded perceptions are stored in memory, potentially affecting the assessment of future events (Augoustinos & Walker, 1995). The combination of newly encoded input along with old knowledge in memory provides the basis for further processing, resulting in inferences and judgements (Srull, 1983).
Cosmides (1989, cited in Fiedler & Bless, 2001) approach to logical reasoning illustrates the importance of the social component in human thinking/intelligence. Cosmides (1989, cited in Fiedler & Bless 2001) claims that failure in logical reasoning tasks can be attributed to the fact that they are usually detached from the social context in which reasoning ability has evolved. This is illustrated with the Wasons (1966, cited in Wiggins, Wiggins & Vander Zanden , 1994) card problem.
Cosmides (1989, cited in Eysenck & Keane 2000) reports people have no difficulty solving logically equivalent problems, when the rule consists of a social contract, reminiscent of social exchange principles. Similarly, Schaller’s (1992, cited in Baron & Byrne, 1997) work on statistical reasoning highlights how social involvement can trigger logical thinking, suggesting failure to complete logical reasoning tasks can be attributed to a lack of motivation and social involvement, which in turn affects the way new tasks/information is perceived.
Smith and DeCoster (2000), claim that the two distinct processing modes draw on the memory systems in fundamentally different ways. Associative processing (top-down) operates pre-consciously and is based on the properties of the slow learning system (Bargh, 1994). For example when an individual meets a women, the targets gender may elicit retrieval of stored gender stereotypes. Using currently available cues to retrieve representations stored on previous occasions where similar cues were present, information can fill in unobserved details and change the way individuals perceive existing features.
Conversely, the defining feature of rule-based (bottom-up) processing is that it uses symbolically represented and intentionally accessed/retrieved knowledge to guide processing (Bargh, 1994). Rule based processing is more effortful and time consuming than associative processing, requiring cognitive capacity as well as motivation. Because rule based processing requires attention, it is subject to distraction and disruption (Logan, 1988). The associative system therefore deals with responses that are made quickly or when the perceiver is busy or distracted.
Given adequate time, Chaiken, Liberman and Eagly (1989) suggested rule based responses may override associative responses. Hogg and Vaughn (1998) explain that schemas can influence the encoding of new information. Existing cognitive structures (schemas), can ‘fill in’ data that’s missing from incoming social information. Schemas can search for the relevant information to complete the stimulus, or they can fill missing values with default information. Schematic processing can lead to biased judgements, and provide short-cuts when processing information by the use of heuristics (Chen & Chaiken, 1999).
With limited information people use the representativeness heuristic (Kahneman and Tversky 1972, 1973, cited in Pennington, Gillen & Hill, 1999) which determines to what degree a specific stimulus is representative of a more general category, causing triggered judgements. Baron and Byrne (1997) explain that mental processing is often subjected to information overload, the ability to process information is exceeded by the amount of information available.
Strategies known as heuristics, (‘rules of thumb’) are used to make complex inferences simpler by streamlining the available information. The use of heuristics, similarly to schematic processing, can lead to errors being made. Tversky and Kahneman (1973, cited in Augoustinos & Walker, 1995) identified the three most commonly used heuristics as; ‘representativeness’, ‘availability’ and ‘anchoring and adjustment’. The representativeness heuristic is where judgements are made (re; a person, object or event) based on how similar or representative it is of a category or prototype.
Tversky and Kahneman (1973, cited in Pennington, Gillen & Hill, 1999) demonstrated that when given a description of a man, representative of an engineer, and told that the man had a 30% or 70% chance of being an engineer, participants ignored these base rates seemingly basing their judgements on the personal description, using top-down processing, relying on stereotypical information to make assumptions. Those participants not given a personal description judged the likelihood in accordance with the base rate, employing bottom-up processing.
Tversky and Kahneman (1982, cited in Pennington, Gillen & Hill , 1999) provided testable evidence for the availability heuristic, showing that although there are more words in the English language that have ‘k’ as the third rather than the first letter, participants claim the opposite to be true. These words are easier to think of, and more available from memory. The availability heuristic can be closely aligned with the principle of priming. A process where stimuli that heightens the availability of certain categories causes them to come more readily to mind.
The anchoring and adjustment heuristic is used as an anchor base from which decisions/inferences are made. Greenberg, Williams and O’Brien (1986) demonstrated this by asking a mock jury to consider the harshest possible sentence before making their verdict, this jury gave harsher sentences than another mock jury instructed to consider the most lenient sentence, suggesting each jury based their decision on the first (harshest/lenient) sentence considered, making only small adjustments. Drawing on the research of Petty and Cacioppo (1981) and Chaiken (1980), Chen and Chaiken (1999) suggest heuristic processing is the default processing mode.
People only go beyond heuristic processing when they need to; be accurate, defend an attitude, or create a positive impression; and have enough time and cognitive capacity for more effortful processing. To understand the role of heuristics & schemas an understanding of how information is stored and categorised, and how the encoding of such information can be affected by the process of priming must be understood. Hogg and Vaughan (1988) explain instances are more or less typical in terms of a range of attributes, with a most typical or prototypical instance representing the category.
Prototypes are defined as cognitive representations of the category, standards against which familiar resemblances are assessed and category membership decided. Exemplars are also used to categorise individuals. Brewer (1988) suggests as people become more familiar with a category, they shift from using prototypical to exemplar representations. How external stimuli is perceived and internalised is dependant on what categories are accessible at the time of the perception. Recently activated thoughts influence the way new information is perceived and organised, which biases the way we organise and interpret subsequent information.
Bargh (1994) showed that thinking about a certain trait construct, makes the construct/s more accessible. Priming is a way of recalling accessible categories or schemas that we already have in mind and can influence interpretation of following stimuli (Herr, 1986). Any priming influence must therefore be related to top-down processing. In many studies (Srull & Wyer, 1979, cited in Bargh, 1996; Hansen & Hansen, 1988, cited in Lord, 1997) of construct accessibility and priming, subsequent thoughts are found to be closely assimilated with previously activated thoughts.
Srull and Wyer (1979, cited in Bargh, 1996) showed that when primed with hostile words in a word comprehension test, participants viewed ‘Donalds’ behaviour to be more hostile (than assertive) compared with participants primed with neutral words. Stack, Schwarz, Bless, Kubler & Wanke (1993, cited in Hogg & Vaughn, 1998) found that when people are made aware of the prime, they over-adjust away from their initial reaction, displaying contrast rather than assimilation. When circumstances prime an extreme version of a trait construct, the result is a rebound or contrast effect.
Herr (1986) demonstrated this by repeating Srull & Wyer’s (1979, cited in Bargh, 1996) study. This time participants were primed with the names; Hitler and Charles Manson. These participants then judged ‘Donalds’ behaviour as assertive rather than hostile. This could be because by comparison to the anchor point (Hitler, Charles Manson) ‘Donalds’ behaviour was rather benign. Pryor and Ostrom (1981, cited in Hogg & Vaughn, 1998) showed that person perception information can be stored in a number of ways.
People can be clustered under attributes or groups or by their individual traits, behaviours or appearances. Social memory can therefore be organised by person or by group. Sedikides and Ostrom (1988, cited in Hogg & Vaughn, 1998) suggest the preferred mode of organisation is by person, which produces more accurate, easily recalled memories. Organisation by person is more likely to occur when it concerns real people, whom we expect to interact with at a later time (Srull, 1983).
Neuberg and Fiske (1987) demonstrated this, showing perceivers categorized target people as former mental patients even when provided with rich information about the individual. However, if the perceiver expected to interact with the target person, they paid more attention to the individuating information. Upon first encounters with strangers, organisation by group membership is likely to occur – people will be pigeonholed, and stored in terms of stereotypic attributes of a salient social category (Hogg & Vaughn, 1998).
Over time organisation may shift from group to individual (Duck, 1977). Processing and interpretation will shift from top-down to bottom-up as we start to get to know the individual and make decisions based on their individual behaviour rather than attributing them to a pre-determined social group. In the area of person-perception, there is considerable evidence (Fiske, 1998, cited in Smith, Marques, Miller, & Mackie, 2003) that the default mode of processing, with the absence of either motivation or capacity, is top-down, this produces stereotyping.
Motivation and capacity appear to increase the perceiver’s use of individuating information (Neuberg & Fiske, 1987). Smith et al (2003) therefore predict that previous exposure to information about the target individual should decrease the perceivers processing of individuating information, thus increasing the perceivers use of heuristic processing and stereotypes. In real-life relationships previous encounters coupled with depth of friendship both motivate and enable individual processing, thereby reducing stereotyping.
Smith et al (2003) have used two experiments to test the hypothesis that repeated exposure increases the impact on social judgments of social group stereotypes, relative to individuating information. In the first experiment participants were shown forty photos, then later shown twelve with personal descriptions (of this twelve, six were within the original forty). As predicted, those descriptions paired with previously seen photos were rated more stereotypically than the novel photos, despite the presence of counter stereotypic behavioural information.
These results were shown to stand true, even when alterations were made to account for the findings of Zajonc’s (1968, cited in Smith, Marques, Miller, & Mackie, 2003) work which showed exposure leads to increased liking for familiar objects. In experiment two, participants read information about a criminal case (involving assault) and received information about the defendants category membership, which was either stereotypically related to aggressiveness or not. Participants also learned specific items of evidence that suggested the defendant was or was not guilty.
In this second experiment it was tested whether repeating items of evidence – the individuating information itself – would increase stereotyping. In the no previous exposure condition – a skinhead was judged to be approximately as guilty as a priest. Conversely, when participants had been previously exposed to some of the information, the skinhead was judged as guiltier than the priest. It was concluded that previous exposure reduced the impact of the repeated information on judgements, and also increased the impact of stereotypes.
Previous exposure led participants to rely more on the stereotype and less on evidence, even though it was the evidence itself that was repeated. Both experiments show that familiarity acts as a cue to generally reducing analytic processing of entire familiar objects – not just that the familiar subset of the information is given less analytic processing. Hastie (1980, cited in Lord, 1997) found inconsistent or unexpected information is recalled better than information consistent with expectation.
Fiedler and Bless (2001) suggest that by forming an overall impression of the situation – in order to make sense of the inconsistent information – the individual invests greater cognitive effort (bottom-up processing). When stimulus behaviours refer to groups rather than individuals, the enhanced memory for inconsistent information is lost (Stern, Marrs, Millar and Cole (1984, cited in Fielder & Bless, 2001). Fielder and Bless (2001) interpret Stern et als (1984) results as showing that less cognitive effort is needed to reconcile inconsistencies between members of a group than within the same individual.
Conversely, Stangor & McMillan (1992) suggest consistent information has a recall advantage as it is derivable from world knowledge. Top-down processing can be relied on to pull information from past experience – using less cognitive effort than bottom-up processing. It can be argued that the inconsistency advantage reflects the bottom-up processing of discrepant stimuli, whilst the enhanced memory of consistent information reflects the top-down influence of embedding knowledge structures.
Question direction and wording can bias processing (Swann, Giuliano & Wegner, 1982, cited in Taylor, Peplau & Sears, 1994). People look for confirming rather than disconfirming evidence, as shown in the Wason task. Snyder & Swann (1978, cited in Smith & Mackie 2000) demonstrated that participants ask questions which provide confirmatory rather than disconfirming evidence to the trait in mind. Semin & Stack (1980, cited in Fielder & Bless, 2001) demonstrated that even completely arbitrary hypotheses, separate from the individual’s beliefs and expectations, can be verified through social actions.
The strength of verification bias is highlighted in a study by Wegner, Kerker and Beattie (1981, cited in Fielder & Bless, 2001) who found that even innuendo entailing denial, such as ‘Bob Talbert not linked to Mafia’ encourages the mental act of imagining a scene where Bob Talbert is linked to the Mafia, this constructive experience may leave a trace in memory, which can confound with observed information, making it difficult to erase the innuendo effect. Even when using deeper cognitive processes (bottom-up) our prior knowledge (top-down) may still influence our thoughts and decisions.
Judgement biases can reflect the distributions of stimuli in the social world prior to any cognitive distortions or motivational forces. Hamilton and Gifford (1976, cited in Lord, 1997) constructed a series of desirable and undesirable behaviours by two groups (A and B). Group A had eighteen desirable and eight undesirable, whilst group B had nine desirable and four undesirable (positivity was uncorrelated within the groups). Observations about group A are more frequent, making A the majority and B the minority, the stimulus distribution is skewed.
A more negative impression is created of group B, the minority group, and there is tendency to associate positive behaviours with the majority group (A). This phenomenon can be related to the permanent source of minority derogation (Fiedler & Bless, 2001). People appear more sensitive to frequency and less sensitive to ratio. It can therefore be argued that this is not a cognitive bias, but rather a predetermined outcome influenced by the frequency distribution of stimuli in the environment.
A danger of top-down processing is illusory correlation. An overestimation of the relationship between two distinct variables can strengthen a stereotype, making it highly resistant to change. Fielder, Hemmeter and Hoffman (1984, cited in Fiedler & Bless, 2001) demonstrated an illusory correlation in an experiment where students were erroneously reported to have proposed more liberal educational attitudes than conservative clerks.
Fiedler & Bless (2001) suggest Fielder et als (1984) results along with similar findings (Hamilton and Rose, 1980 cited in Fiedler & Bless, 2001) show the biased frequency judgements were correlated with a corresponding recall bias. They also discuss that the results show the participants perception and comprehension of the attitude statements. Participants have relied on past experience and expectations to draw conclusions, overlooking the actual stimuli, relying heavily on top-down processing.
Mental capacity, motivation and exposure are all factors which have been empirically tested and shown to influence which processing strategy will be implemented (either bottom-up or top-down). Even when thoughtful Bottom-up processing is engaged, the effects of prior knowledge are still very prominent and there effect cannot be ignored. In our everyday lives, our experiences coupled with recently learnt information is nearly always expected to benefit us in the future, it can therefore be argued that our social experiences have taught us to relay heavily on top-down processing.
The central finding of the empirical studies discussed here has been that social behaviour can be triggered by features of the environment. Behavioural responses can occur in the absence of any conscious involvement or intervention and the danger of this is increased stereotyping which can result in prejudices. The realisation that cognition is very much embedded within a social system rather than merely the individual brain places social psychology at the very core of cognitive sciences.