How to make a semantic differential. Experimental study of pragmatic meanings using the semantic differential method

As a result of mastering the material in the chapter, the student must:

know

• theoretical and practical foundations of the design and use of the semantic differential (SD) method;
• how the SD form defines the research space;
• how the test subject works with the SD form;

be able to

• use SD forms;
• compare DM profiles;
• identify semantic universals of group evaluation and interpret them;
• identify significant differences in group scores on each scale and interpret them;
• identify group assessment factors and interpret them;
• identify group assessment clusters and interpret them;
• draw up specialized SDs for the research task and use them;
• compare estimates for all of the above processing methods and interpret them;

own

• methods of selection, preparation and use of SD forms;
• methods of primary processing of data obtained using SD;
• methods of mathematical processing of matrices;
• methods of analysis, interpretation and synthesis of data obtained using SD.

Standard semantic differential

The subject is asked to evaluate the stimulus (meaning) using the proposed set of antonym scales.

light heavy

fast – slow

active – passive

strong – weak

good bad

The set of scales shown in the example is bipolar(literally - bipolar, the subject chooses one of two possible poles

assessments). More common as variants of diabetes scaled sets of antonyms, when working with which the subject evaluates the degree of expression of a particular property (quality) in a stimulus.

Please rate your idea of ​​the drink offered to you as follows. Here is a list of adjectives grouped in pairs, expressing qualitatively opposite characteristics of the concept being evaluated. Circle the number (from the series 3210123) that, in your opinion, most accurately determines the degree of expression of this particular quality (characteristic) of a given drink, provided that 0 means the quality is not expressed; 1 – weakly expressed; 2 – moderately expressed; 3 – strongly expressed.

(For the full version of this CD with complete instructions, see Appendix 13).

The possibility of a formalized description of semantic differences between stimuli (the possibility of differentiation) determined the name of the technique - semantic differential (SD).

SD is a modified subjective scaling procedure. Procedures similar to the use of SD are often referred to in the literature as multidimensional scaling procedures(each scale is a dimension, many rating scales (space), many dimensions (as degrees of freedom), multidimensional representation of results). When using SD and other multidimensional scaling procedures, it is assumed that all scale scores are independent of each other (the number of degrees of freedom of assessment coincides with the number of scales), but there is no evidence for this assumption yet. If each rating scale is considered as a dimension of the value assessment space, then SD defines a multidimensional value assessment space (stimulus description), which is called semantic space (SP).

The semantic space defined by the standard SD (see Appendix 13) has three integrating factors: evaluation, strength, activity. This space, based on the abbreviation of the first letters of factors in the literature, is called OCA space(assessment – ​​strength – activity) or EPA space (evaluation – potency – activity). According to the data of C. Osgood (Osgood, 1980) and his colleagues, repeatedly confirmed by various researchers, integrating factors are universal (invariant) in relation to the language of the subjects and correspond to the three-component model of describing emotions identified by W. Wundt (Wundt, 1912) (pleasure - tension - excitement).

In the 21-scale LSD (see Appendix 12), seven scales (1, 4, 7, 11, etc. - every third) allow you to evaluate the stimulus (yourself, colleague, wife, boss, cat, etc.) by factor "assessment", seven scales (2,5,8, etc. - every third) - for the factor "strength" and, accordingly, seven scales (3, 6, 9, etc.) - for the factor " activity". These are the factors of the form (questionnaire) laid down by the compiler of the form. They should not be confused with outcome factors (outcome factor structure).

As initial presentation of results experiments using SD are used two-dimensional(one dimension of the table is the SD scale; the second is the subjects) or three-dimensional matrices(third dimension – list of stimuli) in which the assessment results are recorded. Sometimes three-dimensional tables are called slang data cubes, which can mislead students, since these tables often have the shape of a parallelepiped.

50 subjects (first dimension of the table) evaluate 10 stimuli (second dimension) using a 45-scale SD (third dimension). Such a table (matrix) of data (50 × 10 × 45) cannot be called a cube.

These matrices are usually completed in the format of common statistical programs. When filling out the primary result matrices based on the SD forms filled out by the test subjects: 1) for bipolar SD, the subject’s choice of the left pole of a pair of antonyms (scale) is designated by zero, the right pole by one; 2) for scaled LEDs - numbers to the left of zero are written with a minus sign, and numbers to the right of zero are written with a plus sign.

Sometimes scales from 0 to 7 or from 0 to 5 points are used.

Such scales make it difficult for subjects to work, since instead of expressing two qualities, they have to give a differentiated assessment of the quality of the right side of the scale. In addition, the system of marking during training (the more, the better) leaves its mark on the independence of assessment on SD scales, interweaving the parameter (worse is better) into the assessment.

A two-dimensional matrix of results is used when assessing one stimulus by a group of subjects (one dimension is scales, the second is subjects) or when assessing a set of stimuli by one subject (one dimension is scales, the second is stimuli). A three-dimensional matrix is ​​used when assessing a set of stimuli by a group of subjects (the third dimension is stimuli).

Another meaning of the term "semantic space"" is determined by the fact that the evaluated stimulus receives the subject's assessment on each of the scales (dimensions), which makes it possible to describe the stimulus as a point or vector in a given multidimensional SP, to distinguish between assessments of stimuli made by different subjects, and to describe their differences as differences of points or vectors in a multidimensional SP. When working with a group of subjects, we obtain a set of estimates (points or vectors in the SP specified by factors). This allows us to calculate the distance between points in space and angles (cosines of angles) between vectors (from the origin to the point). To evaluate the stimulus, data reduction (integration) methods are used: Based on the assessment of the significance of the frequency of choice of a certain pole of each scale, the method of semantic universals is based, cluster analysis is based on the assessment of distances between points in the SP, and factor analysis is based on the assessment of the angles between vectors.

Factors group stimulus ratings set, although with some loss of analyzed data, a new SP of evaluation criteria, as a rule, less dimensional, but more integrative than the space specified by SD scales (for more details, see paragraph 6.4). The possibility of obtaining a complex multidimensional assessment of a stimulus in SP and the relative complexity of processing procedures allows some researchers to assume that the properties of factor SP are similar to some properties of consciousness, and thus consider SP operational models of the structure of individual consciousness.

This name is very promising, but, in addition to the already repeatedly described limitations of mathematical modeling of the properties of consciousness, one should point out the fact that mathematical processing of assessment results is possible only if many (group) subjects or (very rarely) one The subject evaluates a variety of stimuli. Therefore, such models are not models of individual consciousness. Moreover, there is an unlimited number of different SDs and their modifications (or rather, limited only by the researcher’s ability to come up with “relevant” scales, as it seems to him, for the subject area being studied). This fact makes such “models” simply incomparable. Attributing mathematical properties of SP to consciousness (attributing to reality the properties of a mathematical model) is a methodological error.

With some caution, working with SP allows us to talk specifically about modeling the evaluation of a stimulus in a multidimensional evaluation space specified by the researcher. If (an assumption not yet proven in psychosemantics) SD scales mainly cover parameters of stimulus evaluation that are significant for subjects (stimulus values), then the description of the evaluation (universals, factors, clusters, etc.) allows modeling, for example, in conjunction S – ABOUT – R(in psychosemantics: stimulus – meaning – response (action)), assessment, but not yet a reaction at all and, moreover, not an action of the subject. Therefore, the results of psychosemantic assessment (universals, factor and cluster structures) more often are used to compare values, but Almost never as models of consciousness for predicting activity.

Similarly, operational models of the structure of consciousness could be called the result of processing a stimulus assessment using the method of semantic universals (SP assessments include only scales that are significant for assessment), and cluster structures for describing assessments (SP includes parameters for generalizing stimulus assessments on scales).

To partially level out the shortcomings and evaluate the stimuli of a certain subject area, experimenters often create special cialized diabetes(see paragraph 6.10). Specialized SD (subject specific) are called denotative, in contrast to wide-profile LEDs, called connotative.

Full processing of results experiment using SD includes:

• 1) comparison of assessment profiles;
• 2) identifying group universals of assessment;
• 3) qualitative analysis of assessment universals;
• 4) comparison and qualitative analysis of the comparison of evaluation universals of different stimuli or different subjects (groups of subjects) using SD;
• 5) determination of significant differences in group assessment for each DM scale, their discussion;
• 6) identifying the factor structure of the assessment;
• 7) qualitative analysis of the factor structure;
• 8) comparison and qualitative analysis of the comparison of the factor structure of the assessment of different stimuli or different subjects (groups of subjects);
• 9) identifying the cluster structure of the assessment;
• 10) qualitative analysis of the cluster structure;
• 11) comparison and qualitative analysis of the comparison of the cluster structure of evaluation of different stimuli or different subjects (groups of subjects);
• 12) qualitative comparison of universals, factor and cluster structure of assessment.

Only antonym adjectives are not necessarily used as rating scales. Currently, unipolar SDs (see Appendix 20), verbal SDs (see Appendix 22) and SDs based on other parts of speech have been developed. Based on the assumption of a possible categorical system of figurative information and its connection with linguistic categorical V.F. Petrenko and his colleagues (Petrenko, 1983, etc.) attempted to construct non-verbal SD (visual, etc.), but the procedures for processing and interpreting the data obtained using non-verbal SDs, have not yet been standardized (see Appendix 23).

Advantages SD - the ability to evaluate any stimuli, compactness, the possibility of working with large groups of subjects, the ability to standardize results and procedures for comparing the results of different subjects and groups of subjects, removing speech stamps using scales specified by the experimenter.

Flaws SD – limited possible set of rating scales, the possibility of having rating scales that are insignificant for the subject and the absence of rating scales that are significant for the subject.

• The terms used in the list are explained below.

The semantic differential (SD) is one of the projective methods of sociology, based on the achievements of psychosemantics, and was developed by a group of American psychologists led by Charles Osgood in 1952. It is used in studies related to human perception and behavior, with the analysis of social attitudes and personal meanings. The SD method is a combination of the controlled association method and scaling procedures.

Psychosemantic methods transfer information from the cognitive level (and the research task is always formulated in its terms) to the affective level, where this information is encoded not by linguistic forms, but by various sensations.

The method of semantic differential is based on the phenomenon of synesthesia (thinking by analogy, when some sensory perceptions arise under the influence of others) and is an operational way of “capturing” the emotional side of the meaning perceived by an individual in objects. SD makes it possible to identify unconscious associative connections between objects in people’s minds.

The SD method makes it possible to find a system of latent factors within the framework of which an individual evaluates objects. Essentially, semantic space is a research model of the structure of individual consciousness, and the task is to determine where the object being studied is located in this space.

The tested objects (name, brand, packaging, etc.) are assessed on a number of bimodal seven-point scales, the poles of which are usually specified verbally using antonyms: good - bad, warm - cold, active - passive, etc. It is assumed that a person is able to evaluate the object being studied by correlating the intensity of internal experience about the object with a given rating scale. The divisions of the scale record different degrees of a given quality of an object. Scales that correlate with each other are grouped into independent factors that form a semantic space.

Along with verbal ones, non-verbal semantic differentials have also been developed, where graphic oppositions, paintings, and photographic portraits are used as poles of scales.

Research often uses monopolar scales, with the help of which objects are assessed according to the severity of one property: how good the object is, how warm it is, etc. In the case of bimodal scales, the respondent evaluates where object “A” is located for him on the “expensive - cheap” scale, and with unimodal scales, he evaluates how much the property “expensiveness” is inherent in object “A”. The use of unimodal scales is due to the fact that often antonymous adjectives are not actually complete opposites - bad is not always bad.

In the classic version of Charles Osgood, exclusively connotative features were used as scales, which reflected not the objective properties of the assessed object or concept, but the subjectively significant aspects of the object or concept for the respondent.

In marketing research, a recognized tool for studying the image of a corporation, brand or product is denotative scales, which do not always consist only of antonym adjectives, but are, as a rule, phrases, phrases that express expectations, product characteristics, both negative and positive. Similar products from different manufacturing companies can be evaluated on a “worth the money” scale, for example, banks - by the level of reliability, profitability, etc.

To preserve the “spirit” of the method and capture the still affective elements of the attitude, a set of scales (15-25 scales) is used. The result of the technique is not directly calculated average values ​​of objects on each of the scales, but latent factors obtained during a special analysis procedure, on the basis of which the semantic space of perception of objects is formed and a map of their relative positions is constructed. It is important to select a sufficient number of scales and test them on experts or conduct an association experiment on representatives of the target group in order to avoid the danger of researcher subjectivity when choosing scales.

SD scales do not describe reality, but are a metaphorical expression of the subject’s states and relationships (the instructions that respondents receive call for: “When making assessments, be guided by your own feelings, not knowledge”). In the resulting space of affective meanings, a convergence of concepts to which a person reacts in a similar way and a separation of concepts that have different emotional backgrounds are recorded. The distance between concepts is expressed by a certain number, which in general allows one to distinguish between assessments of: a) the same concept by different individuals (or different groups); b) different concepts by the same individual (or group); c) one concept by the same individual (or group) at different times.

The number of identified factors corresponds to the structure of the emotional perception of a given class of objects, so, for example, when evaluating a bank, only two factors can be identified: reliability and profitability, while a car can be assessed according to the criteria of “fashionability, style,” “prestige, status,” “ prices”, “operational efficiency”, “after-sales service networks”, etc.

The procedure for developing a semantic differential methodology within a specific research project usually consists of the following steps:

Formation and testing of a list of adjectives, statements to describe the objects being tested (names, concepts, types of packaging, brands, etc.). The level of awareness at which the respondent will evaluate the measured object depends on the selected features. By focusing on denotative scales, we expand the semantic space, increasing information about objects and inevitably losing information about subjects, which is not so critical in marketing research.

Mathematical processing of the resulting data matrix: object - respondent - scale. Typically, a factor analysis procedure is used, which allows one to identify latent assessment criteria into which the initial scales are added. It is important to note that to obtain significant results, relatively small samples - 30-50 people - are sufficient, due to the fact that the unit of analysis is not the respondent, but the ratings he gives to objects. Considering that each of the 30-50 respondents evaluates 7-10 objects on 15-25 scales, the total sample size turns out to be quite sufficient to draw statistically significant conclusions.

Placement of evaluated objects in the constructed semantic space, analysis of the resulting distribution. Estimation of the distance between the tested objects and the ideal object (for example, ideal yogurt, car, “myself,” etc.), to determine the “positive” poles of the factors. For example, if we received the factor “fashionability, stylishness, brightness” of a car, then it is important to understand whether the high ratings of our brand on this factor are positive for the target audience or not. Perhaps the ideal car for them is a reliable, conservative “iron horse”, economical in fuel consumption and without any special quirks in design.

Stage 1 Formation and testing of a list of statements.

The tools used in the semantic differential technique usually consist of a table of the following type: scales are placed in the rows, and assessed objects are placed in the columns. The instructions offered to the respondent are formulated approximately as follows: “Please rate the characteristics of each of the brands “…” on a scale from 0 to 5, where 0 means no such property, and 5 means the property is expressed to the maximum. In the “ideal ...” column, write down what properties good ... should have, using a scale from 0 to 5, where 0 - such a property should not exist, and 5 - the property should be inherent in the product to the maximum extent.”

Considering that a completely sufficient sample for a homogeneous group of respondents within the framework of this methodology is 30-50 people, it is often convenient to collect information in parallel with a focus group study. Considering that usually the launch of a new brand, name, or packaging is accompanied by a series of focus groups, then during 3-5 groups it is possible to collect 30-50 questionnaires. This sample size turns out to be quite sufficient to supplement the conscious, rational information provided by respondents with assessments of the affective elements of the attitude, i.e. collect extraconscious, emotional, irrational data that the semantic differential technique allows you to obtain.

Figure 4 - Example of a semantic differential table for object evaluation

Stage 2. Mathematical processing of results and their interpretation

The SD technique makes it possible to quite clearly process the results and interpret them using the simplest statistical characteristics. The average value of the measured value, standard deviation, and correlation coefficient are proposed as such characteristics. Primary processing of the results consists of compiling a statistical series of the measured value for each object under study. Then the average statistical value of the measured value for the sample and a measure of the unanimity of estimates, expressed by the standard deviation, are calculated. Once the average ratings of each object for the three measured indicators have been identified, it is interesting to trace their interdependence. Thus, the algorithm for mathematical processing of SD results is as follows:

Step 1. Drawing up a statistical series in the form of a table.

X i - assessment of a certain quality of an object on a seven-point scale;

n i - frequency of the value X i, i.e. how many times was the score X i given when assessing the object according to the parameter under study by all respondents in the aggregate.

Step 2. Calculation of the average value.

If K respondents participated in the survey, then the average value is calculated using the formula:

n=M*K, since the quality under study is assessed K by respondents in the developed form M times (in M pairs of antonymous adjectives). The average value of X serves as an indicator of the overall assessment of a given quality of an object by the entire class, being at the same time a fairly objective characteristic, since it allows one to neutralize the influence of subjective factors (for example, the bias of individual respondents in relation to a given object at the time of the survey).

Step 3. Calculate the standard deviation.

The standard deviation serves as an indicator of the measure of dispersion of the values ​​of a quantity around its average value X, i.e. measures of unanimity and cohesion of respondents in assessing a given quality of an object. The standard deviation is calculated as the square root of the variance y x = vD x, where the variance D x, in turn, is calculated by the formula:

The described three steps of mathematical processing of diagnostic data reveal the picture of the respondents’ perception of the objects under study. This allows you to clearly present the results of the analysis.

The data obtained after the above-described processing can be compared with each other by calculating their correlation. This stage of processing aims to establish to what extent the respondents’ attitude towards the object is related to its individual characteristics.

Step 4. Calculation of the correlation of the obtained estimates.

When determining the correlation coefficient, firstly, the average value of the ratings of each of the indicators for all assessed objects is calculated. Let's say a respondent evaluates n objects. Based on activity, the 1st object was assessed by the average value of A j. Then the average score of indicator A of all objects:

Average P indicator score:

Then the correlation coefficient of A and P r A,P:

(covariance); , - standard deviations of the values ​​A j and O j from their average values, which are found as follows:

As a result of calculating the correlation of ratings, one can clearly see the psychological mechanism for constructing the relationship of respondents’ ratings to the objects under study.

Stage 3. Presentation of the location of the tested brands in the semantic space.

After the mathematical processing stage, several main factors can be identified and the location of the tested brands in the semantic space formed by the identified latent factors can be presented.

The results ultimately turn out to be quite clear and quite easy to interpret: the figure shows that the ideal product should have high quality and an affordable price (for the sake of clarity of the example, fairly obvious properties were selected). In terms of quality factor, brands 1 and 2 are closest to the ideal product, and in terms of price factor, brands 4 and 5. By assessing the set of criteria, we can conclude that brand 1 is closest to the ideal.

In a similar way, you can test, for example, name options, choosing names that evoke the most positive emotions, while being associated with a specific product and evoking an image and association with the corresponding valuable qualities.

Interesting results can be obtained by comparing products that do not compete with each other, but have a similar basis, which makes comparison possible and helps to identify new positively assessed qualities of a product or brand and transfer them to a new product field (invention for use).

For example, an assessment of plastic cards in general, in order to understand what properties of fuel plastic cards need to be developed, and the use of which would help in capturing the fuel card market.

The semantic differential technique allows, when studying a brand, to identify an emotional attitude towards it (the affective component of the attitude), not burdened by rationalizing motives (the cognitive aspect). Identify how the potential consumer feels about the brand, i.e. predict his real behavior, and not words about actions.

The semantic differential allows you to draw statistically significant conclusions on small samples (sufficient material can be collected from 3-5 homogeneous focus groups) due to the fact that the unit of analysis is not the respondent, but the assessment (on average, each respondent evaluates 7-10 objects of 15 -25 scales, i.e. gives 100-250 ratings).

The SD method allows us to identify the structure of latent factors, criteria on the basis of which respondents construct evaluations of various brands. Accordingly, using the SD method, it is possible to construct a map of the placement of brands of interest in the structure of factors, thereby obtaining a visual, relatively easily interpreted research result.

The use of an “ideal” object in the SD methodology, along with the brands being tested, allows us to determine the desired directions of development, possible threats to the brand, and the most significant (although sometimes unconscious by the consumer) properties of the product.

The use of SD methodology in marketing research allows one to evaluate a brand and its elements (name, packaging, corporate identity, etc.), obtaining statistically significant estimates of the deep structures of consumers’ consciousness in the course of a relatively inexpensive and small-scale study.

PROJECTIVE METHODS

The psychosemantic method as a method of measurement and as a method of analysis. Semantic differential by C. Osgood. The method of unfinished sentences as an approach to the study of social expectations and social stereotypes. The Twenty “I” Test as a method for studying personal self-identification.

The index analysis and ranking procedure will demonstrate that it is sometimes difficult to classify the “method” of sociological research as either a measurement method or an analysis method. It all depends on the context, the research situation and from the goal to achieve which method is applied. The absence of a coherent conceptual apparatus in sociology is an internal specificity and property of our science. Therefore, some methods are ambiguous:

1) they act as measurement techniques,

2) in the role of analysis methods.

Each such procedure involves a specific technique for collecting information and specific mathematical processing of empirical data. Therefore the concept diagnostic procedure in contrast to the concept of method and technique is more acceptable. Unfortunately, this concept is not widely used in the sociological literature.

This includes psychological procedures, borrowed by sociologists from psychology. These procedures can be called tests .

Some tests seem to measure personal characteristics, others - group characteristics. There are a lot of tests. Let us consider the procedures related to the so-called projective methods.

When studying social reality, the researcher asks the question of how this very reality is perceived by specific people, refracted in their minds, and transformed into certain social norms and images. To penetrate into the inner world of a person, it is impossible to use predetermined schemes, unambiguously interpreted concepts and categories. Accordingly, strictly formalized, structured methods of collecting information do not work in this case. There is a need to attract concepts and categories that people themselves use to organize their own everyday life experience.

These methods are based, as it were, on the projection of the subjective properties of a person onto the screen of a sociologist. The sociologist offers the respondent stimuli (signs, texts, pictures, situations) and, based on the reactions, determines hidden, unconscious thought processes, needs, images, etc.

Let's remember the logical square. It is to some extent a projective technique. By asking the respondent two questions, we identified his subjective inclinations and desires in situations that were unrealistic for him. According to these personal meanings determined the meaning: the degree of satisfaction with studies as the strength of motivation.

As an example, consider three projective methods:

1) Ch. Osgood's method of semantic differential(Ch. Osgood), is used in sociology to solve a wide range of problems and is interpreted both as a method of measurement and as a method of analysis.

2) unfinished sentence method¾ is used in sociology as an approach to the study of social expectations and social norms.

3) the twenty “I” test ¾ is used to study personal self-identification. In the last two cases we are talking about working with textual information, which also determined the choice of these particular methods.

C. Osgood's semantic differential (SDO)

This method was developed in the mid-50s by Charles Osgood to study people’s emotional attitudes towards certain concepts in order to determine their meaning. It is well described in the literature (for example, a review can be found in the work). The LMS method is as follows. The respondent is asked to express his attitude towards a certain object (concept, image) using a set of bipolar scales, mostly seven-point, for example, such as:

Extreme positions on the scales are described by verbal antonyms. The set of scales forms the initial space scales The number of gradations on the scale may be less than seven. Extreme positions can also be non-verbal. For example, C. Osgood used the signs “black circle ¾ white circle”, “up arrow ¾ down arrow” etc. when studying representatives of different linguistic cultures (Indians, Mexicans, Japanese and Americans) in their attitude to different concepts.

Why were the scales listed above selected? Various experiments with different sets of bipolar scales gave the same result. The entire set of scales seemed to fall into three main groups, into three factors named as Strength, Activity, Attitude .

This phenomenon was discovered by Charles Osgood and called synesthesia. For our purposes, all that matters is how it manifests itself. Factor¾ a group of scales on which the object’s ratings are similar. Of the above seven scales, the “strength” factor includes scales 1 and 2 (weak ¾ strong, male ¾ female), the “activity” factor scale 3¾4 (active ¾ passive, slow ¾ fast), the “attitude” factor 5¾7 (usual ¾ unusual, false ¾ true, good ¾ bad). One can argue in a similar way for the case of the rest of the 20 classical scales introduced by Charles Osgood. In addition to the seven pairs listed, these twenty include the following pairs:

cruel ¾ kind,

curve ¾ straight,

loose ¾ punctual,

tasty ¾ tasteless,

unsuccessful ¾ successful,

hard ¾ soft,

stupid ¾ smart,

new ¾ old,

unimportant ¾ important,

sharp ¾ rounded,

cold-blooded ¾ enthusiastic,

colorless ¾ colorful,

unusual ¾ ordinary;

beautiful ¾ ugly.

These scales were selected after a huge number of experiments with different pairs of adjectives through the use of mathematical methods of factor analysis (you don't need to think about that yet). It was for this set of scales that their three-factor structure was proven. In other words, we, as respondents, evaluate any object (concept, image) presented to us emotionally, mainly according to three factors or in three-dimensional space. This space is called semantic space . For this reason, the term “semantic” is used in the name of the method. Images of objects in semantic space occupy a very specific specific position. By analyzing the location of objects, one can draw conclusions about the proximity of images. For an analogy, let us recall school geometry and the concepts of “two-dimensional space” (it is simply called a plane) and “three-dimensional space”.

The idea of ​​the SDO method was further developed. In actual research, a sociologist has three options when working with this approach to the study of social reality, or three strategies for forming a set of initial scales:

a) use well-known, seemingly classical, scales with slight adjustments;

b) take advantage of the results of the work of other researchers;

c) try to form your own semantic space of factors.

In the first case adjustment is required to exclude pairs that are incorrect when assessing specific objects. Let us give an example of one problem in which the LMS method was used. The task is to conduct a typological analysis of the perception of the images of 10 popular politicians, i.e., to identify different groups of politicians. At the same time, the image of politicians assigned to the same group is the same. The objects of the study were teachers of the sociology department of one of the universities. The study was purely methodological in nature, so there was no problem of representativeness and the sample was small. A classic set of scales was used, with the exception of pairs such as bad ¾ good, smart ¾ stupid, etc. Such pairs were excluded due to their excessive specificity and incorrectness for the case of our task.

Second strategy, namely the use of the results of other researchers, is possible in the following situation. Let's assume that we are talking about mass surveys and the sociologist does not have the opportunity to experiment with a large number of scales and form his own semantic space. Then, to select scales for his research, he proceeds as follows. For example, in the task of studying the images of politicians, we chose the same number of scales for each of the three factors. And we learned from the literature about which factor and with what weight the scale relates.

Third strategy¾ the formation of one’s own semantic space for evaluating objects arises in deep analytical studies, when SDS is the main approach to the study of social reality. Then the sociologist initially forms a set of scales that are specific and not associative in nature (like the 20 classic ones). In this case, he is obliged to test the hypothesis about the existence of factors and find out what they are and how many there are. This is necessary to move on to the analysis of images in the space of these factors. Each problem, for a set of objects specified by the researcher, can have its own factor space with a different number of factors.

Regardless of the chosen strategy for using the LMS method, the analysis of object images is carried out in the following way. Let's talk about it using the example of studying the images of political leaders. Having received the initial data ¾ of the assessment of each teacher for each of the ten political leaders, ¾ calculated average rating for each of the three factors for each object (leader).

The score for a factor for an object is equal to the sum of scores for all scales included in this factor and for all respondents, divided by a value equal to the product of the number of scales and the number of respondents.

In this case, instead of writing a simple formula to calculate the average score, we verbally described its content. As you can see, this is inconvenient. This is why a sociologist needs mathematics, because it gives comfortable description language.

After calculating the average score for all factors (strength, activity, attitude) separately, the transition to calculating the proximity between political leaders is carried out. This can be done in three-dimensional space, or in two-dimensional space (by choosing two factors that are important to you from three).

The second case is simpler. Moreover, from school geometry you remember how the distance between points (political leaders) on a plane is calculated (two selected factors form it). Let's still use the formula. Let us denote by D (A,B) distance between two political leaders A And IN in the space of factors, through d i (A,B)¾ difference in average grades i-mu, factor. The number of factors is equal to k. Then about the proximity of the images of political leaders A And IN can be judged by the meaning of the so-called differential.

Having carried out calculations for all pairs, and there will be 45 of them for ten political leaders, we obtain the so-called proximity matrix or a matrix of the “object ¾ object” type. Let's remember Thurstone's method of paired comparisons. There were proximity matrices there too, only of a different nature.

If anyone is interested in our research, it should be noted that four types of images of political leaders have emerged. The study was conducted in March 1996. The image of President Boris Yeltsin was different from all other images. The situation with V. Zhirinovsky is similar. The images of V. Chernomyrdin, A. Lebed, G. Zyuganov were close. All other political leaders formed the fourth group. The interpretation of this result is not important to us. This problem is of interest to us only as an example illustrating the use of the LMS method in mass surveys.

2.Method of unfinished sentences (INP)

An example of a study that was also methodological in nature. It was associated with the problems of students’ subjective perception of the image "cultured person" The task was to determine the meaning that people put into the concept they use ¾ “cultured person”. You can come across this expression quite often in everyday speech. What do people put into it? What kind of person do they call cultured and what are the criteria for culture? What is the image of a “cultured person” in their minds? Are there different types (types) of perception of this image? If they exist, what are they? What is the semantic space of image perception and what dimension is it?

It is absolutely clear that attempts to discover a clearly formulated understanding of the image in the respondent will be futile, since in people’s minds this image has blurry, amorphous outlines.

Application MNE gives the greatest effect precisely in those studies where there is a need to identify a person’s subjective perception of social reality, personal meanings individuals, stereotypes, images, standards, value orientations of people, etc.

Respondents are presented with a set of unfinished sentences and asked to write their completion. The verbally expressed reactions of respondents to the first part of the sentences constitute the basis on which the main characteristics of the image being studied can be identified. We can say that these reactions carry information about the norms, values, stereotypes, standards, and images that exist in society and have been internalized by the individual. Their reconstruction is carried out on the basis of information collected during the survey. Below are just some of the open-ended sentences we used in this study.

1) What distinguishes a cultured person from other people...

2) A cultured person must...

3) Usually cultured people...

4) Closest to the concept of “cultured person” is the concept...

5) All cultural people are united by...

6) I call a cultured person...

7) The most important thing for a cultured person...

8) A cultured person never...

9) A cultured person always...

10) A cultured person has...

11) The opposite of a cultured person...

12) I cannot call that person a cultured person...

In these proposals, respondents retain the opportunity to choose the criteria themselves and determine the semantic basis of the answer; the respondents’ reactions are not limited to predetermined options. In general, we can count on the fact that the developed proposals make it possible to obtain respondent reactions that are minimally distorted by the influence of the researcher. The subject is forced to speak in his own words, as a result of which, upon completion, he uses the categories that he uses in everyday life to organize his life experience.

It was assumed that when composing endings to sentences, respondents should have used different justifications. In the broad sense of the word justification ¾ this is a moral prescription, culturally defined stereotypical patterns of goals, motives, etc. In a narrow sense, it is convenient to use the concept "elementary justifications" to denote semantic particles that are obtained as if by division text (sentence endings) into separate indivisible parts.

At the first stage of analysis, the sentence completion texts for all respondents are divided into elementary justifications. After this, justifications that are similar in meaning are grouped, resulting in the formation of separate elements , which express one or another characteristic of the image.

For example, justifications like: « What distinguishes a cultured person from other people...": “...style of speech”, “...ability to clearly express one’s thoughts”; “Usually cultured people...”:“...they know how to explain their point of view without resorting to shouting”, “...they don’t shout at others, they don’t swear”¾obviously have a common semantic focus and refer to one element.

Accordingly, they differ from justifications of the type: “What distinguishes a cultured person from other people is ...": "...presence of principles", "...will", "...high moral level"; "Usually cultured people...”: “... do not hesitate in making moral decisions.” These justifications also constitute a separate element,

The first element marks the manner of communication of a “cultured person.” It can be conditionally called "Speech and Thoughts" . The second ¾ are moral principles and the inner world of a “cultured person”, and it can be called conditionally "Inner world" . Some elements undergo generalization at an even higher level, thereby forming concepts of a higher level of abstraction. Isolation of elementary justifications, and then elements, is something other than logical formalization when analyzing texts. The next stage of the analysis consists of comparing the image of a cultured person for different respondents. For example, based on the frequency of occurrence elementary justifications And elements . But here it is impossible to do without quantitative calculations, without mathematical formalization . To do this, complex information coding is first required.

Due to the fact that for now we are mainly interested in measurement problems, we present a fragment related to this problem. As noted, the sentences “The concept “closest to the concept of a “cultured person” is...” and “The opposite of a cultured person...” make it possible to establish concepts similar to and opposite to the image being studied. Most of the answers represent the same images of what a “cultured person” is. Thus, respondents named the closest to this concept: “intelligent person” ¾ 37%, “well-mannered person” ¾ 16%, “polite” ¾ 11% and “educated” ¾ 9%. Opposite concepts: “boor” ¾ 28%, “uncultured person” ¾ 13%, ignoramus ¾ 8%. By themselves, these data are not very informative. Nevertheless, one can raise the question of constructing a “cultured person” – “uncultured person” scale. You can also consider concepts similar to these poles and, using, for example, the semantic differential method, evaluate the similarity of all these concepts.

In addition, of interest are those respondents who contrast “cultured people” with certain social groups, among which homeless people and criminals stand out. It can be assumed that in the minds of these people, certain social groups have a certain measure of culture, so it would be of great interest, using the same semantic techniques, to determine such a measure of culture in different social strata of the population. From this we conclude that within the same study, various methods, approaches to obtaining and analyzing information can and should be used simultaneously.

The peculiarity of the data obtained using the method of unfinished sentences is the possibility of repeated access to primary information. In this case, various bases for classifying text information are used to solve various research problems. The MNE method is successfully used to study social expectations and social identity. MNP is also used in mass surveys as a measurement method. Only in this case one or more sentences are used.

Test of twenty “I” (TDY)

This method was developed in the 50s by M. Kuhn and T. McPartland to study the image of one’s own "I", for the study of self-determination or personal identification. The method of collecting information is quite simple. The respondent is offered a sheet of paper with the heading “Who Am I”, with twenty numbered lines. They turn to him with a request to answer, as to himself, “Who I" and write down your answers quickly enough; preferably in the form of nouns. In the order in which they come to mind. There is no need to worry about logic or the importance of the answers.

The authors of this technique found that The answers are divided into four classes. Two of them relate to objective self-determinations (let’s call them K1 and K2), and the other two are subjective (let’s call them KZ and K4).

To class K1 include such individual self-definitions as a “physical” object (I am a man, I am a woman).

Class K2 constitute self-definitions that represent the individual as a social object (I am a member of society, I am a student, I am a music lover, I am a scientist, I am a teacher).

To the class of subjective self-determinations KZ include those that are associated with socially relevant characteristics of behavior (I am ¾ a useless person, I am ¾ a pessimist, I am ¾ lucky, I like to listen to music, I like to drink in good company).

Class K4 form those self-definitions that are, to one degree or another, irrelevant in relation to social behavior, and also irrelevant to the task of self-identity posed by the test (to live ¾ to die).

Below are real data ¾ these are the answers of three students from a study we conducted among future linguists. Try to interpret these data, because these three situations will be encountered in research.

Three students' answers

It must be borne in mind that not all respondents give a complete answer. The number of completed lines in itself characterizes the personality of the respondent. As a rule, the respondent often does not have enough nouns to answer, and some have an “excess” of them. Respondents have or do not have self-definitions from the four indicated above. Based on this, we can move on to a formal “description” of the respondent. Assign to each a set consisting of zeros and ones.

The theoretically possible 16 sets are listed below:

0000 0001 0010 0011 0100 0101 0110 0111

1000 1001 1010 1011 1100 1101 1110 1111

If the respondent has self-definitions from all four classes, then he is assigned set 1111. If the respondent has self-definitions of only class K2, then it is put in correspondence 0100. The number of possible sets is equal to 2 4 =16.

In practice, not all sets are found. This coding of textual information allows us to identify separate groups of respondents with the same self-identification structure. Thus, the sociologist finds typological groups, typological syndromes for the study of self-identification. Other ways of formalizing textual information obtained through the twenty self-determination test are also possible.

What conclusions can be drawn from such a superficial examination of methods? SDO, MNP, twenty self-determination test.

1. All three ¾ methodological procedures, in which it makes no sense to separate the collection technique, measurement technique, and analysis technique. At the same time, in real research they can be used as a technique for measuring the properties of social objects, a technique for analyzing social reality, and a technique for collecting empirical data.

2. Each type of text information requires a specific logical formalization for analysis. Only after this comes the turn for mathematical formalization.

3. All three can be used for different purposes and, in particular, for carrying out a typological analysis of social phenomena.

1. Measurement is a component of information analysis associated with the procedure for obtaining data initial for analysis. Certain techniques for studying social reality are called both measurement techniques and analysis techniques (logical and analytical indices, ranking). Some measurement techniques (C. Osgood's semantic differential, the method of unfinished sentences) are also interpreted as approaches to the analysis of social reality. Measurement begins with a model for studying the properties of social phenomena.

2. Dimension ¾ is scaling (unidimensional or multidimensional). Measurement ¾ is the procedure for obtaining a scale (Likert scale, Thurstone scale, Gutgmann scale). Measurement ¾ obtaining the scale itself, i.e., a ruler with gradations (assuming the existence of a one-dimensional scale). Measurement ¾ is a diagnostic procedure.

3. If we correlate measurement techniques with types of empirical data, we obtain the following conclusions. In the first type we are talking about the metric level of measurement and the problem of measurement is reduced mainly to the formation of analytical indices and ranking. In the second type, measurement occurs as the coding of verbal judgments or the use of graphic scales. And finally, the problem of measurement is inscribed, absorbed, conditioned by the different “origins” of texts.

©2015-2019 site
All rights belong to their authors. This site does not claim authorship, but provides free use.
Page creation date: 2017-10-25

Semantic differential method- this is one of the objective methods of constructing subjective semantic spaces(cm. ). Belongs to the method class experimental semantics(cm. ). The concept of subjective semantic space is an operational model of representation categorical structure individual consciousness as mathematical space, the coordinate axes of which correspond to the foundations implicitly inherent in the individual categorization(see), and the values ​​of some content area are specified as coordinate points or vectors located in this space. Based on the model of the categorical structure of individual consciousness, the distinction, classification and evaluation of any objects, concepts, and so on is carried out by analyzing their meanings, since the placement of certain meanings in the semantic space allows for their analysis.

The semantic differential method was developed in 1955 by a group of American psychologists led by Charles Osgood during the study of the mechanisms of synesthesia and has been widely used in studies related to human perception and behavior, with the analysis of social attitudes and personal meanings. The semantic differential method is used in psychology, sociology, decision theory, mass communications theory, advertising and a number of other areas. This method is also used in the study of cognitive (cognitive) aspects of consciousness and self-awareness of an individual. As Osgood believes, it allows one to measure so-called connotative meanings, that is, those states that follow the perception of a stimulus symbol and necessarily precede meaningful operations with the symbol. The closest analogue of connotative meaning from the apparatus of Soviet psychology is the concept personal meaning- “meanings for the subject” (A. N. Leontiev, A. A. Leontiev).

The semantic differential method is a combination of the controlled association method and scaling procedures. The measured objects (concepts, images, individual characters and others) are assessed on a number of bipolar gradual (three-, five-, seven-point) scales, the poles of which are specified using verbal antonyms. Concept assessments on individual scales correlate with each other, and with the help of factor analysis it is possible to identify bundles of such highly correlated scales and group them into factors. From the content side, a factor can be considered as a semantic invariant of the content of the scales included in the factor. Grouping scales into factors allows us to move from describing objects using features specified by scales to a more capacious description of objects using a smaller set of categories-factors, presenting the content of an object, its connotative meaning, as a set of factors given with different weight coefficients. In the geometric representation of the semantic space, the categories-factors act as the coordinate axes of a certain n-dimensional semantic space (where the dimension of space is determined by the number of independent, non-correlated factors), and the connotative meanings of the analyzed content area are specified as coordinate points or vectors within this space.

Semantic space is a kind of metalanguage for describing meanings, which allows, by dividing their contents in a fixed alphabet of categories-factors, to carry out a semantic analysis of these meanings, to make judgments about their similarities and differences by calculating the distances between the coordinate points in space corresponding to the values. In Osgood's works, three main factors were identified: “Evaluation”, “Strength”, “Activity”, combining many scales, and Cartesian three-dimensional space was used to differentiate connotative meanings. Working with a large number of scales and concepts, American researchers A. Lavoie and P. Bentler managed to expand the set of basic factors and, along with the factors “Evaluation”, “Strength”, “Activity”, identified the factors “Complexity”, “Order”, “ Reality", "Ordinary". Similar results were obtained by V. F. Petrenko using the material of Russian vocabulary.

Along with universal semantic differentials built on the basis of vocabulary from various semantic classes, particular semantic differentials are also constructed for limited conceptual classes. For example, a number of private semantic spaces have been constructed: “personal semantic differential”, structuring adjectives - personality properties; “differential of political terms”, and so on. The construction of such private semantic spaces allows for a more refined semantic analysis, and the factor structures themselves can be interpreted as a categorical grid of a given conceptual class. Private semantic spaces built for a given social population or individual do not have cross-cultural invariance and carry differential psychological characteristics. The latter makes it possible to use them in the psychology of individual differences, but requires a procedure for constructing a subjective semantic space for each individual study. Along with verbal semantic differentials, nonverbal semantic differentials(Nonverbal Semantic Differential), using graphic oppositions, paintings and photographic portraits as scales.

In addition to the denotative meaning, each object has a connotative meaning for a person. This meaning is personal, gained through direct experience. If one fine sunny day a heavy apple fell on my head, I lost consciousness, and when I woke up, I realized that I was lying in a pile of cow dung, then for the rest of my life I will stay away from clusters of large apples on trees. In the example given, the experience of “communicating” with an apple is very vivid. Usually the connotative meaning is more hidden.
I will give other examples of connotative meanings. The rector of a university can be assessed by his students as a firm and cold person. This does not mean that the rector’s body density and temperature differs from the average of other people outside normal limits.
In other words, connotative meaning is an emotion towards the object being evaluated.
What does semantics have to do with it? We introduce the definition according to Tolstoy. Semantics is a branch of linguistics and logic that studies the problems of meaning, meaning and interpretation of signs and symbolic expressions. Accordingly, psychosemantics is the study of a person’s psychological perception of the meanings and meanings of various kinds of objects. Psychosemantics includes such methods as semantic differential, repertory lattices, etc.
The task of psychosematics is very interesting - the construction of a semantic space J. I.e. system of latent factors within which a person works. Why did you walk around the puddle in front of the entrance on the right side this morning, although it was more convenient on the left?
Why does sociology need SD? For example, a sociologist might try to identify types of people with similar perceptions of objects. If the object is the advertised product, then for each individual type it is more effective to make a separate advertisement with the desired perception J
The big advantage of SD is that, using “hard” methods, it provides information about the subtle psychological structures of a person’s perception of objects.

Semantic differential technique

What did Osgood suggest? The emotion of the meaning of a concept will be revealed if a person points to the position of the concept in question in the system of connotative features. Those. will indicate the location of the object in the “emotional” coordinate system. For example, evaluate a political leader: is he warm or cold, fluffy or prickly?
So, let’s prepare several pairs of emotions (connotative features). Pairs naturally contain opposite emotional colors: sweet and sour, black and white, good and evil. Each pair contains several gradations. If you want to use factor analysis in your analysis, you need data defined by an interval scale. To do this, there must be seven gradations (the more gradations, the more your scale shifts from ordinal to interval type).

Osgood found that in most cases, any of the connotative pairs hides one of three possible options: strength, evaluation (attitude), activity. In other words, if we take an object, let respondents evaluate it based on hundreds of similar pairs, and then conduct a cluster analysis of all these pairs, we will see that all pairs are divided into three groups: strength, evaluation, activity. Those. When we perceive any object of reality, we “attribute points” to this object according to three characteristics: strength (strong-weak), assessment (bad-good) and activity (fast-slow).