NOTE NEW DUE DATE FOR BRIEF PROJECT DESCRIPTION. THANK YOU!
OVERVIEW
COURSE PAPER TOPIC STATEMENT DUE October 2 

WHAT'S NEEDED FOR THE PROJECT UPDATE IN OCTOBER IS ALSO ON THE COURSE PAPER PAGE (just scroll down)

THEORIES OF SOCIAL PSYCHOLOGY
SUSAN CAROL LOSH

SYLLABUS

SYP 5105-01           FALL 2017

GUIDE TO THE MATERIAL: FOUR
AFFECT AND ATTITUDES

 

KEY TAKEAWAYS:
  • Evaluation--or positive or negative affect--is the core that distinguishes attitudes from beliefs or opinions
  • Emotion generally signifies more arousal than evaluation (although emotion is affect, too)
  • Attitudes are feelings about targets; they include dimensions of direction (+ or -), accessibility, centrality, salience, extremity, complexity 
  • Attitudes differ from opinions (more belief) or values (more general and enduring)
  • Attitudes can--and should be--measured several different ways, and both verbally and nonverbally. Question format can introduce many biases
  • For example, Likert Scales (degree of agreement or disagreement) are particularly subject to acquiescence response set or extremity response set
    • **Not convinced? See the Public Opinion Quarterly article (Tourangeau et al) OPTIONAL under Guide 4 for examples of great care with survey question construction and experimentation--does it make a difference? WHEN does it make a difference?**
  • The research literature on attitudes and predicted behavior is mixed. Voting and consumer behavior often demonstrate high correspondence
  • Both voting and consumer behavior are relatively specific, planned, time-limited and often private attitudes and behaviors
  • Choice and commitment and situational norms are crucial to predict attitude-action consistency
  • Emotions and their display may be influenced by the social situation, cultural norms and labels from others

 
GUIDE 1: INTRODUCTION
GUIDE 2: ISSUES IN METHODS
GUIDE 3: A SOCIAL PERCEPTION PRIMER
GUIDE 4: AFFECT AND ATTITUDES
GUIDE 5: PERSONALITY AND THE SELF
GUIDE 6: LEARNING THEORIES AND SOCIALIZATION
GUIDE 7: AN INTRODUCTION TO GROUPS
GUIDE 8: GROUP STRUCTURE & INFLUENCE

 
 ATTITUDES
MEASURING ATTITUDES
ATTITUDES & BEHAVIOR
ATTITUDE CHANGE
PUBLIC OPINION

Like it? Hate it? Research with the semantic differential (see below) indicates that evaluation--or positive or negative affect--is the core that distinguishes attitudes from beliefs or opinions. Don't confuse affect or emotion (I hate orange drinks) with beliefs or cognitions (that drink is orange) or actions (I will pour the orange drink down the sink drain.) The cognitive (the section we just finished, information oriented), the affective (attitudes), and the conative (action orientation)  obviously interrelate but are different entities.

It's also important to recognize that when many social psychologists discuss "affect" they basically mean evaluation--like, dislike, positive or negative. Although evaluation obviously has emotional overtones, it is still somewhat distinct from emotions, which cover a much broader range--joy, guilt, shame and so forth. Emotions denote an arousal component which may or may not be present with evaluation.

Work on social perception dovetails nicely with the cognitive orientation that has dominated social and behavioral research for the past 50 years. The attempt has been made to study attitudes, or affective orientations toward objects, in much the same way. This cognitive approach has, in fact, carried us some distance in the study of attitudes, but it does leave largely untouched the messier, untidy businesses of love and lust, and jealousy and hate. In fact, even the topic of emotional intelligence has been "cognicized" as it has developed. You may want to consider whether we need to add or amplify cognitive and reinforcement approaches to include emotions, or whether totally new dimensions may be needed.

ATTITUDES

Attitudes are feelings--or evaluations--about targets. These vary in:

"Opinions" are more cognitive beliefs about a target or object. Typically opinions are presented as "factual" statements.

Values concern broad goals  (e.g., freedom) and can represent primitive [basic] beliefs. Attitudes are more specific (e.g., supporting voting rights).

World views are clusters of organized, interrelated, possibly hierarchical attitudes. "Liberal" or "conservative" self-definitions are examples of world views and allow us to predict many specific attitudes.

It is a bad idea to use concepts such as opinions, attitudes, or values as though they were interchangeable. They are not interchangeable.
I have know some students and even professionals to do so. However, this confounding can make attitudes and perhaps subsequent actions more difficult to understand.

For example, consider two individuals with attitudes about capital punishment ("the death penalty"):

Like beliefs, attitudes can be formed through second-hand knowledge and are influenced by people they like and respect, and, alternatively, people who are disliked (a boomerang effect, "whatever they are for, I am against").

Attitudes can serve functions for the individual. Four functional areas that were delineated by Daniel Katz several decades ago, which are still quite useful include:

People with very extreme attitudes often have what are called "narrow latitudes of acceptance" and "larger latitudes of rejection," that is, they find fewer attitude positions on a topic acceptable to them than people who hold less extreme positions (these are social judgement theory terms). To some degree, this means that they will be more resistant to persuasion to change their attitudes. On the other hand, such individuals may be more credulous about information that supports their attitudes (confirmation biases) and accept it more unhesitantly (related to the heuristic "affirming the consequent").
 


MEASURING ATTITUDES

Attitudes can--and should be--measured several different ways, and both verbally and nonverbally.

VERY IMPORTANT NOTE: Unless an attitude scale has been validated in diverse ways, INCLUDING BEHAVIORAL MEASURES IF APPROPRIATE, it is inappropriate to (a) use it as a stand-alone measure, claiming the attitude is "whatever the scale measures" and (b) "validate" it by showing correlations with another attitude measure. Attitude measures easily could be correlated because of response format--and NOT because of content. For example, Likert items (degrees of agreement or disagreement) tend to correlate due to acquiescent response set. (See the OPTIONAL recent Tourangeau article on methodology in Public Opinion Quarterly.)

Attitude questions can be "open-ended," that is, the response is like an essay in the individual's own words. Or questionnaire items can be "closed," or similar to multiple-choice exams where the respondent is provided with a pre-existing set of response categories and asked to select one response category. Researchers tend to like to use lots of closed questions because they are quick to administer, and are much easier to score and place into a computer processing format. However, if your survey includes many items that have the same or a very similar response format, please be advised that your results could reflect some type of format response set instead of the content that you envisioned (see below for a section on response format effects).

Long sets of questions, in which you verbally (or in writing) ask questions and record the answers are basically surveys, whether you call them a "standardized test," a "personality inventory," an "assessment," or something else entirely. As such, your test, assessment, or inventory is subject to all the same problems, by and large, that public opinion surveys face (and maybe more, because public opinion surveys typically take good, representative samples and hence have good external validity.)

One common attitude measure is the Likert scale. Respondents are given a statement, typically an evaluative statement of some kind ("I was shocked by the destruction wrought by Hurricane Irma"), then asked to indicate their degree of agreement or disagreement with that statement. Researchers love Likert scales because they can be completed very rapidly (a skilled interviewer can administer 5 LIKERT ITEMS PER MINUTE; yes, this is real), thus providing large amounts of data in a very short time, and they are easily and unambiguously coded. In contrast, open-ended responses are difficult, ambiguous, and time-consuming (possibly expensive) to code, despite the richness and multidimensionality of responses.
 
 

 
DID YOU KNOW?

The technical term "Likert Scale" refers ONLY to indicating degree of agreement or disagreement with a statement. This type of measure was developed by Rensis Likert in the 1920s. Keep in mind that one purpose of graduate school is to give you a common vocabulary with other professionals in your discipline: it is incorrect to refer to any kind of graduated, ordinal scale (e.g., "Very true of me, somewhat true of me, a little true of me, not at all true of me) as a Likert Scale.

Likert Scales are particularly subject to acquiescence response set or extremity response set (see below) where respondents agree or disagree in a pattern, instead of responding to individual item content. Incidently, some research has found that using more than five responses to a Likert scale (basically strongly agree to strongly disagree) does not improve the reliability or validity of the scale, although it does take people more time to complete the measure.

In  semantic differential scales, the respondent is typically provided with polar opposite pairs of adjectives ("happy-sad" "simple-complex") and asked to describe the attitude object along continua for each pair of adjectives. Researchers have found the evaluative dimension tends to dominate these descriptions (although a "soft-hard" dimension also frequently emerges). Semantic differential scales can be subject to extremity response set, or the tendency to rate a target at polar ends versus in the middle of the scale.

Projective tests do not ask about the attitude object directly. Instead, you may be given a picture related to the object, an ambiguous picture (an inkblot), or an incomplete sentence. Your task is to tell a story about the picture or complete the sentence. Projective tests are sometimes used if the material is considered preconscious, unconscious, or subconscious, and the subject is thought to arouse some type of internal psychological conflict. The idea is that by using a picture or vague sentence, you will project your feelings into the story or sentence completion where they can be scored for imagery. One example is the Murray Thematic Apperception Test. However, both the scoring and the inferences made through using projective tests are subject to considerable scholarly controversy. In particular, the practice of inferring an individual's personality through the use of projective testing has been subject to criticism. Among other things, so-called   "projective tests" often reflect an individual's culture rather than nonconscious feelings. On the other hand, projective tests have been found to be useful to help identify physical or sexual abuse in children.

My students and I scored children's drawings of scientists, veterinarians, and teachers. We investigated hypotheses that these drawings (in the case of scientists) reflect children's feeling about scientists (e.g., whether they can visualize themselves in that career). However, the gender ambiguity in many of the boys' drawings, in particular, led us to be skeptical about the value of these particular "projective tests" such as drawings in young children.

Indirect methods observe behavior. For example, if you find a "lost letter," researchers theorize that you are more likely to return it if the group addressed holds attitudes consistent with your own attitudes. Thus, it is supposed to indirectly measure public opinion on controversial topics. For example, you might estimate attitudes toward abortion by using letters addressed to different  groups (e.g., Planned Parenthood versus a Right to Life group). Another indirect way would be to study interruption patterns in mixed-sex groups. Yet another popular example is "horn-honking". Researchers place different bumper stickers on their cars (for example, for Democratic or Republican presidential candidates), then stall the car at an intersection, recording the number and ferocity of honking horns. The problem, of course, is that all of these behaviors reflect all kinds of tendencies (such as absent-mindedly sticking the letter in your purse or pocket, and meaning to mail it...) besides the attitude you really want to measure, and this can introduce enormous error variance into attitude estimates. Some of the same issues on inferences raised about the use of projective testing apply here, too.
 


ATTITUDES AND BEHAVIOR

Do attitudes do a good job of predicting behavior? While diverse attitudes are interesting to study, many researchers believe that the true value of these studies lies in predicting how people will behave. Sort of, "the wish is parent to the deed."

Very early research on attitudes and behavior produced dismal and depressing results. One of the most famous early studies was undertaken by sociologist Richard LaPiere, who travelled the United States in the 1930s with a Chinese couple, stopping at many of the small, locally owned motels that were then prevalent. LaPiere and his friends always were able to rent a room. After he returned home, LaPiere wrote these same motels. In his letter, he mentioned "by the way, my friends are Chinese." Nearly half the motels never responded. (LaPiere counted a nonresponse as a refusal of the Chinese couple.) The others wrote back regretting that a room was unavailable. LaPiere concluded that different sets of social forces influenced attitudes and behaviors, and attitudes were a poor predictor of behaviors. Although there are several methodological critques about LaPiere's work (see below), historically his study is very important because it was one of the first to use field testing to compare behavior and attitudes. The results were so interesting to researchers that there were many partial replications of LaPiere's study.

Research undertaken through the 1950s tended to agree with LaPiere's. Sizable numbers of studies centered around White American attitudes toward Black, Asian, or Hispanic Americans. Restaurants admitted Black customers, prejudiced customers were waited on by Asian or Hispanic American department store clerks and didn't even seem to notice the ethnicity of counter clerks.

Starting in the 1960s and 1970s, this research became more multidimensional and sophisticated. As a result, we have considerably more knowledge of how attitudes and behavior relate.

For one thing, there are two GIGANTIC areas where attitude questionnaires do an excellent job of predicting behavior: voting and consumer studies. (The 2001 AAPOR t-shirt slogan, in a parody of the 2000 Presidential election reads: "polling, now more accurate than the election itself.") As long as representative samples are taken with solid response rates, estimates of national and state votes (or, indeed, any other pending election in which probability samples of at least 1500 are taken) are generally accurate within a few percentage points. The same is typically true for predicting large consumer purchases such as automobiles.
 
 
 
Because most presidential election polls in 2016 predicted a win for Hillary Clinton, President Trump's win was considered by many (including many pollsters, for whom election polls are "demonstrations of accuracy" to obtain business from TV stations or toothpaste manufacturers) to indicate the polls were flawed. However, that actually was not true. The polls were predicting the national USA popular vote and were very accurate. However, the US president is elected through the Electoral College and its votes are apportioned differently from the national popular vote. State polls vary enormously in quality and that's what was needed to predict the electoral college vote. Although most of the time the electoral college and the national popular vote predict the same winner, there are exceptions. In 2000, Al Gore won the popular vote for president but the electoral college elected George W. Bush president.
 

 
 
 
Telephone surveys using Random Digit Dialing have been one of the most cost effective probability sample ways of gauging the mood of the USA general public since the 1980s. The big problem now with telephone surveys is exceptionally low response rates, e.g., 10% or less. Cell phone only households are now a majority of households and nearly all cell phones have caller identification (as do many landlines). The rise of "robo caller" attempts at phishing and sales, especially those that "spoof" your area code and exchange, have led many people to simply no longer answer their telephone unless they recognize the caller. Very low response rates increasingly resemble "self selected" respondents rather than probability samples. Many government agencies have now switched from telephone to "snail" mail surveys; many companies now use online samples (which have their own representativeness problems.)

What did LaPiere and the early attitude-action researchers "do wrong?" What do some of the political pundits or market researchers "do right?"

Let's look at the LaPiere study first. It is a deserved classic because to our knowledge, Richard LaPiere was the first person to systematically test the attitude-behavior relationship. And I think we always need to keep sight of that. The United States in the 1930s was a very overt ethnically intolerant nation. Many Euro Protestant Christian-Americans didn't like Chinese, Blacks, Jews, Catholics, Hispanics, the Irish, or much of anyone else. ALL these groups were then subject to prejudice and discrimination in the U.S. (and elsewhere, of course). So LaPiere's research decision to look at motels, and with a Chinese couple, was a good choice.

On the other hand, what do voter studies and market surveys have in common? Both voting and purchases are: All these factors help us make good predictions from attitudes to action in both areas.

The more sophisticated research that began in the late 1960s tells us that attitudes can predict behavior if:

IMPORTANT NOTE: Attitudes can be  quite volatile, especially on peripheral issues so people often change their minds on specific, non salient topics. This makes sense. Don't we want people to be open to new information and to reform their attitudes as new information is accessed? This is a "nature of the beast" problem and it is one of the things that makes research on attitudes and values difficult yet so interesting: the target is ever-changing. Although it is irritating how voters misremember who they voted for in presidential elections (the winner was their choice, except for Richard Nixon in 1972), forget about legislation before the ink on the laws is even dry, and think that the topics you want to study are dull. On the other hand, attitudes about important and central people, objects and events tend to be relatively stable over time.
 


ATTITUDE CHANGE

Can "stateways change folkways"? Must we always change attitudes before behavioral change can occur? These aren't just academic questions. Under times of national stress, we may not have the time or leisure to change attitudes one by one and wait for an associated behavior to ensue and change too.

Well, one possibility is propaganda. We send out messages that we hope will change attitudes in "our desired direction." Attitude change researchers believe a central sequence of attention, comprehension and acceptance (a more reasoned approach) or a peripheral, unreasoned process [elaboration likelihood model] occurs. The central sequence of attitude change is analogous to many current approaches in learning theories more generally.

Peripheral processing refers to the expertise, similarity or trustworthiness of persuasive sources rather than actual message content. For example, if you respect and trust former U.S. President, Jimmy Carter, then you will probably adapt many of his current statements as your own attitudes. You may also look to your co-workers or local political leaders. Central processing is more what we consider reasoned action. It may depend on your access to information that you believe is relevant.

Both central and peripheral processing can occur in parallel in the same persuasive attempt.

I will note that the amount of available information, per se, is not necessarily any help. Recent research indicates that many young adults accessing political information on the Internet feel overwhelmed--information overload--with the amount of information that can be accessed.

A critical distinction social psychologists make in processes of persuasion is between compliance and conformity. In compliance, only external behavioral change is desired. External rewards and punishments are often used to ensure compliance.

In  conformity, internally motivated attitude change also occurs. Appeals are often made here to reason, logic, knowledge, self-expression, identification with the source, credibility of the source and so forth. Any attitude acquisition or change that occurs when people feel like they have a choice about whether to accept influence is generally more permanent than when either occurs under compliance. Quite simply, compliance is very costly, to those who create and must enforce it, and for those who accept this influence. Threats and promises influence compliance and can require surveillance.

(These topics will be reviewed in greater depth later this semester under group processes.)

"Immunization effects" can prevent attitude change. If we give you practice at refuting arguments aimed at changing your attitudes, resistance to later appeals rises. [We will also look at "obedience to authority" later.]

Initial research on fear appeals dismissed them as ineffective. People were so turned off and anxious from the message, they never really got to the comprehesion or acceptance part. However, we now know that fear appeals can be effective if they specify easily implemented actions and arouse a moderate amount of fear.

Attitude or attitude-behavior inconsistency can be psychologically and literally physically uncomfortable  and motivate change toward consistency. However, discomfort results only when attitudes are accessible and relevant. Change toward consistency occurs more often if you perceive choice and low environmental pressure, commit to and  feel responsible for your behavior. Balance theory (Heider) is oriented internally toward changing beliefs and attitudes. Cognitive dissonance addresses internal attitude-behavior inconsistency as well as other internal cognitions.
 
 
 
P.S. Cognitive dissonance is about internal individual information processing and change.
Please DO NOT use cognitive dissonance to refer to disagreements across people. This is incorrect.

YES! Behavior can change attitudes, either because:

Problem is: most of us don't typically go out motivated to change our behavior, unless we feel our behavior falls short of culturally prescribed and accepted goals (e.g., it's good to exercise.) So how do we get people outside the laboratory to change their behavior in the first place? The self-perception approach is probably more accurate: many everyday behaviors follow role-related scripts, yield to subtle persuasion or reflect automatic conformity, so we often commit to behaviors where we did not previously hold a corresponding attitude, or our attitude may even contradict our behavior (especially on peripheral issues or in a spontaneous situation). Methods such as the foot-in-the-door technique in which a small request is later followed by a larger one may spark behavior change. Or, motivation for a really desirable reward may spark change.
 
 
A TALE OF A CLASSIC EXPERIMENT

You, a male undergraduate, enter the laboratory and spend an hour at a thoroughly boring task. At last! The experiment's over and you can go home. But not so fast. Two-thirds of the time, the experimenter stops you and asks if you can help. "Normally," he says, "there's someone in the waiting room before you go in who tells you what an interesting experiment this is." "But no one was there" you say. "Precisely," says the experimenter, "Our confederate is home sick today and couldn't come. We would like you to help and tell the next subject what an interesting experiment this is. Naturally we will pay you for your time."

One third of the subjects are offered $20, a truly remarkable sum in the early 1950s when this classic study by Festinger and Carlsmith took place--and not so bad for a few minutes work even now.

The second, random, third are offered a gigantic ONE DOLLAR.

The final random third are not told anything about the missing confederate and simply go home after finishing the task.

The first two-thirds of the male subjects greet an extremely attractive young woman whom they assume is the next subject (she is really the true experimental confederate.) They proceed to tell her how interesting the experiment is.

Then, all the real subjects are given a short questionnaire to assess how much THEY really liked the experiment.

Who liked the experiment the most? Or, actually, who disliked the experiment the least?

Take your guess and then click HERE for the answer of what really happened .

How does reward affect counter-attitudinal behavior?

Dissonance says that smaller rewards cause MORE attitude change, especially in commitment conditions.

Over-sufficient justification makes similar predictions for intrinsically [internally] motivated but not for extrinsically motivated behaviors.

Direct reward or reinforcement effects (more reward=more change) can occur when rewards are expected, not a surprise, not linked to competence, extrinsically motivated, or the person perceives little choice or commitment.
 

THE FIELD OF PUBLIC OPINION

Scholars in public opinion research study many of these same topics. First, they study how people process information, which can relate to stability and change in public opinion. Next, they examine the cognitive underpinnings and how people process information on the influence of question format and response. For example, we know that many people respond to the form rather than to the content of a survey research item. Examples of format response set include acquiescence response set or extremity response set.

One branch of public opinion research is familiar to us from the "pollsters" who ascertain the standing of political candidates and how people respond to public issues.  Another is market research, and may focus on consumer behavior. Much of this research has given us considerable insight on attitude-behavior linkages.

Another part of the public opinion research field studies various forms of propaganda and persuasion campaigns. Scholars in these areas often look at the "face" the media present to the public (e.g., the gender or ethnic composition of television programs) and how media influence public attitudes.

Here is the site for AAPOR, the American Association for Public Opinion Research. AAPOR is about 50-50 academics and public opinion professionals.

CLICK HERE   to access AAPOR.

RESPONSE SETS

In acquiescence response set, people tend to generally agree or disagree with nearly every sweeping attitude statement put before them in a questionnaire.  Well-educated persons have been taught "never say 'always' or 'never,' " so "nay-saying," or disagreement with general attitude statements, rises with education, typically regardless of content. We are less clear about which factors nudge people to gravitate toward extreme responses ("strongly agree" or "strongly disagree") or toward the middle, but we know that this tendency happens too.

The tendency to simply "opt out" and respond "I don't know" is also linked to education. In one recent study about science knowledge in the general population, I found that people who had not graduated high school were three times as likely to give "I don't know" responses as people who had graduate school experience to basic items such as "Does the sun go around the earth or does the earth go around the sun? ". Women were nearly twice as likely as men to say "I don't know." In such cases, you need to reflect on whether a "don't know" response truly represents lack of knowledge or some other factor such as low self-efficacy or the desire to avoid embarrassment. (Also see the optional Tourangeau et al article about "don't knows".)

Any kind of "standardized test" or assessment which contains long sets of items, all with identical response formats, is especially prone to response sets. This is especially true if all items take a Likert-type format. In addition,"personality inventories" which contain long sets of identically phrased items (e.g., "A lot like me," "Somewhat like me," "A little like me," or "Not at all like me") are also subject to response format effects. Unfortunately, a definite, but unknown, specific amount of widely used "reliability coefficients," such as coefficient alpha, is undoubtedly due to the fact that correlations among the items occur because of response set. ...Interesting...but probably not what you hoped for. Coefficient alpha is also influenced by the number of items: larger numbers of items in a "test" produce larger coefficient alphas. However, neither of these factors seem to designate what most of us think of as "reliability".

OPTIONAL: INTERESTED IN DOING A SURVEY or INVENTORY?

OPTIONAL: SEE MORE ON QUESTION CONSTRUCTION HERE

WHAT YOU CAN DO:


Using experimental "split-ballot" techniques, survey researchers also assess whether different forms of a question yield comparable responses and appear to be comparable in meaning. Subsamples of participants from the total survey are selected at random to receive the "same item" with different question wording. We check not only to see whether the univariate frequency distributions are the same, but also whether question wording changes how one concept relates to other variables, such as educational level.
 

THOSE PESKY EMOTIONS

By this time, you have probably noticed that even the attitude research material has a distinct cognitive orientation. That's true. Attempts to measure entities such as "emotional intelligence" (a sensitivity to social situations, tact, humor, etc.) became more successful in terms of publishing when researchers took a more cognitive approach.

The sociological more symbolic interactionist view is that emotions are socially constructed. We learn, often from the outside in, to link internal physical feelings with socially labelled emotions. I am reminded of research on sexuality in film from about 40 years ago ( from The President's Commission on Pornography Report). Researchers showed sexually explicit films to both women and men (in separate rooms). Both showed nearly identical measures of physical arousal. When asked to describe their feelings, the men said they were aroused and the women said they were disgusted! In early theorizing, psychologist Daryl Bem suggested that parents and other older or authority figures take cues for how children feel from their facial expressions, vocal inflections, and bodily postures in the context of the social situation. Then they describe to the children in words how they feel ("you're excited" "you must feel sad"). This line of research is also compatible with Vygotsky's socio-cultural theories of learning.

The social situation also influences not only which emotions are appropriate to display (OK to cry at a wedding but not upon getting a traffic ticket) but (see above) how people actually describe their feelings to themselves.

But attempts to measure what jealousy is, and its effects on behavior, or anger, or joy, are really in their infancy. I think the old cognitive vocabulary is not going to work and we are in the process of developing a new one.



 
OVERVIEW
SYLLABUS

 

This page was built with Netscape Composer.
Susan Carol Losh September 19 2017


Halloween Shelties