�WPC>� � _"� ��/u�S�t��:؞<�el��w]{���p$�`E����b�n�|>#��N7wJ"O�2�9+���^�;R!tL���������8���V"Itƈ@lZ�a���J��a�.����ު�#��V����+���l��pO�e@�j^W9;����jSﳘ�Ȭ��1D���aM��^��.*�?��z���"����!O��Z�M���q����aI��G�d�OiKn ђ�[p�����Rbf�ø�j�b揂:"��[5���߸ev�a�Dg2UÄ�a�r~ |��O�f6�HHP+ mn�e���j� y�9b45Q�q�����0� ���F���2���Q�Gf�����d懾�7L�ޒ�dv� )v��K�2.�Ϸ�bIa ������H@��5�#*�w= c1{P����M�?z]���!'J���f�� #�� �qV��b�'5=��E�2���6��~�;�=jՎq�y �u�YU%:� %� 0d� 0[ 0`w 0]� 0b4 0�� 0�Z 0�! 0�� 0�� 0� 0oC 0�� 0�v 0�: 0�  0�� 0G� 0Z��R 0�? h��E� 0�� BG,pI�� h R! hD\$��+ 0�~, -nX- 0rt.UN�.UN4/4�/��/� j0fJ<aL<�`< B&= 0 �C=��=��>�]? AM#@ 0Kp@ D3�@ D/�@.A 0�KBf�Ba�BUN�B D+=COhC�C��DfXcaZc AYnc��c 0F�d AM�d�Ee 0M fnXff�x��xf��a��5�� h�a�� R1��V��fy�a{���� hrw� DM�� h6�<�6X9`("Courier NewTT&� ' � /Contextual Content Analysis "Don McTavish"0Don McTavish .   h~~��:Default ParaDefault Paragraph Font��XXX��O�<�6X9`("Courier NewTTO��  ��  ��  ��  ��������O�<�6X9`("Courier NewTTO�T�~��:endnote textendnote text�  ���XXX��O�<�6X9`("Courier NewTTO��  ��  ��  ��  ��������O�<�6X9`("Courier NewTTO�^����:endnote refeendnote reference��XXX��O�<�6X9`("Courier NewTTO��  ��  ������  ��  ��������O�<�6X9`("Courier NewTTO�V�~��:footnote texfootnote text�  ���XXX��O�<�6X9`("Courier NewTTO��  ��  ��  ��  ��������O�<�6X9`("Courier NewTTO�`����:footnote reffootnote reference��XXX��O�<�6X9`("Courier NewTTO��  ��  ������  ��  ��������O�<�6X9`("Courier NewTTO�88T��,toc 1toc 1�  �� ��(#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�8;T��,toc 2toc 2�  ����` (#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�8>T��,toc 3toc 3�  ����` � (#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�8>T��,toc 4toc 4�  ����� (#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�8>T��,toc 5toc 5�  ����h(#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�88T��,toc 6toc 6�  �� ��(#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�8!��,toc 7toc 7�  ��  ��  ��  ��  �88T��,toc 8toc 8�  �� ��(#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�88T��,toc 9toc 9�  �� ��(#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�@;T��0index 1index 1�  ����` (#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�@;T��0index 2index 2�  ����` (#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�P5T��8toa headingtoa heading�  �� ��(#��  ��  ��  ��  ��>4�X�` � h�p� x� (#>�@�~��0captioncaption�  ���XXX��O�<�6X9`("Courier NewTTO��  ��  ��  ��  ��������O�<�6X9`("Courier NewTTO�^~~��:_Equation Ca_Equation Caption��XXX��O�<�6X9`("Courier NewTTO��  ��  ��  ��  ��������O�<�6X9`("Courier NewTTO� ��8�)���Fd8���&X��X�&��An�early�version�of�this�paper�was�presented�at�the�Pacific�Sociological�Association�Meetings,� � �Searttle,�April�12,�1984.(f31�$��� �%��!���&�Rf��&���������&�Rf��  ��  �+ +++++++++�TABLE A+(e2��$��� �%��!���&�Rf��&���������&�Rf��  ��  � d f~~~�func�{�\"�Self�\"���``���E�score�}�~�=�~�{0.333�``���``�0.045�}�over�{�0.028�}�~�=�~�+10.29,����"��,�Self���"��x�E��������score�����A� ����0��e�.����333���������0��.�.��S�045��70��J7.��o7028������� ����9 �10��� �.��� �29��� �,9(9999999999999999999999 9 9 9 9 9 9 9 9 9 9 9 999999�TABLE B9@D@@@@@@@@@����MbP?@����Mb�?@����Mbp�@�~j�t�h�@;�O��n�?@@����MbP?@�~j�t�x?@����Mb`�@{�G�zt?@�~j�t��?@@����Mb��@@;�O��n�?@{�G�z��@/�$��?@@y�&1�|�@����MbP?@����Mb�?@{�G�z��@���S㥛?@@y�&1�|�@����Mb��@L7�A`�?@;�O��n��@��~j�t�?@@y�&1�|�@����Mb`�@����Mb�?@y�&1�|�@�O��n�?@@����Mb��@����Mb`?@����Mb�?@{�G�z��@sh��|?�?@ @ ����Mb��@ y�&1�|�@ �� �rh�?@ ��~j�t��@ -����?@ @ y�&1�|�@ y�&1�|�@ �� �rh�?@ {�G�z��@ Zd;�O��?@ @ �@ �@ A@@ 4�@ �TABLE B@ �� �"1�(���&��V��&��������&��V��  ��  �����8(#���X�jX������Contextual�Content�Analysis�󀀀���  �1�  �(f31�$��� �%��!���&�Rf��&���������&�Rf��  ��  �3|x��<�6X9`("Courier NewTT��d�6X@�D�Q�@<�6X9`("Courier NewTTXXx�6X@�D�QX@(J��$�����ˑ�������ˑ���ˑ��  �(�� Z �6Times New Roman Regular(�� Z �6Times New Roman Regular!�1"������ �� ���'���&��V��&��������&��V��  ��  ���X�jX�����  �1�  ��  ���D.G.�McTavish�and�E.B.�Pirro�� � � �81�"���&��V��&��������&��V��  ��  ���;��V������*s�9: ddd Xdd Xdd X��s��,�dd ��,� dd ��,� dd ��+  �� �  �� �Context� ,�dd ��,�Category� ,�dd ��,�Typical�word�or�phrase� -#dd ������-�Traditional� %UU ��%�Guide� %UU ��%�should,�ought,�guard� 3)UU  ������3�� % ��%�Structural�Roles� % ��%�mighty,�military� 3)   ������3�� %��  ��%�Prohibit� %��  ��%�restrict,�watch� 3)��   ������3�� %��  ��%�Ideals� %�� ��%�stability,�honesty� 3)��  ������3�Practical� %YY ��%�Activity� %YY ��%�walk,�buy,�sell� 3)YY  ������3�� % ��%�Merchandise� % ��%�product,�spend� 3)  ������3�� %�� ��%�Strive� %�� ��%�maintenance,�development� 3)��  ������3�� %�� ��%�Organization� %�� ��%�management,�office,�factory,�retail� 3)��  ������3�Emotional� %]] ��%�Happy� %]] ��%�friendly,�wonderful� 3)]]  ������3�� % ��%�Pleasure� %  ��%�gladness,�refreshment� 3)!  ������3�� %��" ��%�Expression�Arena� %��# ��%�museum,�music� 3)��$  ������3�� %� � % ��%�Self�other� %� � & ��%�respond,�wish� 3)� � '  ������3�Analytic� %a a ( ��%�Differentiate� %a a ) ��%�analysis,�analytic� 3)a a *  ������3�� %" " + ��%�Relevant� %" " , ��%�solution,�signify� 3)" " -  ������3�� %� � . ��%�Similarity� %� � / ��%�alike,�comparison� 3)� � 0  ����� �3�� %� � 1  � �%�Scholarly�Nouns� %� � 2  ��%�library,�university,�science�," � � 3   ,��TABLE A�Level 1Level 2Level 3Level 4Level 5��P Pd(f31�$��� �%��!���&�Rf��&���������&�Rf��  ��  ݃Level 1Level 2Level 3Level 4Level 5�Level 1Level 2Level 3Level 4Level 5�Level 1Level 2Level 3Level 4Level 5-/45A<< C��( ��$��� �Figure��  �1�  �� �'��dxd..'�� dxd $� �/1�#���&��V��&��������&��V��  ��  �� �GT����Table��  �1�  ���.���  �Illustrative�Conceptual�Categories�and�Words�More�Likely�To�Be�Found�In�  �Each�of�the�Four�Context�Categories(f31�$��� �%��!���&�Rf��&���������&�Rf��  ��  �TABLE B(�� Z �6Times New Roman Regular'����dxd EE�score�sub�{i,k}�~�=�~�{p�sub�{i,k}�``�``�P�sub�i�}�over�{�S�sub�i�} ��)E���)����)score����i����,����k���)�� �;Y��T�p����i����,����k��c������P��_�i��$�S���<i {� �81�"���&��V��&��������&��V��  ��  ���&��V�����"Work�like�mine�keeps�me�from�doing�my�best."�  ��*s�@A ddd Xdd Xdd X��s��,�dd ��,�dd0��,�dd0��,�dd0��,�dd0��,Hdd0��+  �� � �� � ��� 0�%%  ����0�Accumulation�of�C�score�Deviations� +�%% � �+�� -#%%  ������-��Words� .$�� ����.��Traditional� 5+��3 ����5��Practical� 5+�� 3 ����5��Emotional� 5+�� 3 ����5��Analytic� 5+��3 ����5�Weight�average� =�0#��3  ������=�Work� 1�$�� ����1�+0�.001� 8�+��U ����8�+0�.008� \�O��U �����Mbp�-0.004��������Mbp����\Є0�.004� ��sA��U ����Mbp�-0.004 ����Mbp� ��~j�t�h�-0.003�����~j�t�h�����Є0�.003� �rA��U �~j�t�h�-0.003 �~j�t�h� �;�O��n�?0.009����;�O��n�?����0�.009� ]SF��U ;�O��n�?0.009 ;�O��n�? ������]�like� .$YY ����.�+0�.001� 5+YYU ����5�+0�.006� YOYYU �����Mb`�-0.002��������Mb`����YЄ0�.002� YOAYYU ����Mb`�-0.002 ����Mb`� ����Y�+0�.005� XNYYU ��~j�t��?0.012�����~j�t��?���X�0�.012� ]SFYYU �~j�t��?0.012 �~j�t��? ������]�mine� RH �����Mb��-0.008��������Mb�����RЄ0�.008� xnAU ����Mb��-0.008 ����Mb�� � 0�������x�0�.000� TJ<U  0  ����T�+0�.018� XN U �{�G�z��-0.01����{�G�z�����XЄ0�.010� {q@!U {�G�z��-0.01 {�G�z�� �/�$��?0.021����/�$��?���{�0�.021� ]SF"U /�$��?0.021 /�$��? ������]�keeps� RH��# �y�&1�|�-0.007����y�&1�|����RЄ0�.007� YOA��$U y�&1�|�-0.007 y�&1�|� ����Y�+0�.001� 5+��%U ����5�+0�.016� XN��&U �{�G�z��-0.01����{�G�z�����XЄ0�.010� {q@��'U {�G�z��-0.01 {�G�z�� ����S㥛?0.027�������S㥛?���{�0�.027� ]SF��(U ���S㥛?0.027 ���S㥛? ������]�me� RH��) �y�&1�|�-0.007����y�&1�|����RЄ0�.007� }sA��*U y�&1�|�-0.007 y�&1�|� �����Mb��-0.008��������Mb�����}Є0�.008� YOA��+U ����Mb��-0.008 ����Mb�� ����Y�+0�.033� YO��,U �;�O��n��-0.018����;�O��n�����YЄ0�.018� |rA��-U ;�O��n��-0.018 ;�O��n�� ���~j�t�?0.038������~j�t�?���|�0�.038� ]SF��.U ��~j�t�?0.038 ��~j�t�? ������]�from� RH]]/ �y�&1�|�-0.007����y�&1�|����RЄ0�.007� }sA]]0U y�&1�|�-0.007 y�&1�|� �����Mb`�-0.002��������Mb`����}Є0�.002� YOA]]1U ����Mb`�-0.002 ����Mb`� ����Y�+0�.016� YO]]2U �y�&1�|�-0.007����y�&1�|����YЄ0�.007� |rA]]3U y�&1�|�-0.007 y�&1�|� ��O��n�?0.149�����O��n�?���|�0�.149� ]SF]]4U �O��n�?0.149 �O��n�? ������]�doing� RH5 �����Mb��-0.008��������Mb�����RЄ0�.008� YOA6U ����Mb��-0.008 ����Mb�� ����Y�+0�.002� 5+7U ����5�+0�.016� XN8U �{�G�z��-0.01����{�G�z�����XЄ0�.010� {q@9U {�G�z��-0.01 {�G�z�� �sh��|?�?0.166����sh��|?�?���{�0�.166� ]SF:U sh��|?�?0.166 sh��|?�? ������]�my� RH��; �����Mb��-0.008��������Mb�����RЄ0�.008� }sA��<U ����Mb��-0.008 ����Mb�� �y�&1�|�-0.007����y�&1�|����}Є0�.007� YOA��=U y�&1�|�-0.007 y�&1�|� ����Y�+0�.034� YO��>U ���~j�t��-0.019������~j�t�����YЄ0�.019� |rA��?U ��~j�t��-0.019 ��~j�t�� �-����?0.177����-����?���|�0�.177� d�SF��@U -����?0.177 -����? ������d�best.� Y�H� � A �y�&1�|�-0.007����y�&1�|����YЄ0�.007� ��sA� � BU y�&1�|�-0.007 y�&1�|� �y�&1�|�-0.007����y�&1�|�����Є0�.007� `�OA� � CU y�&1�|�-0.007 y�&1�|� ����`�+0�.034� _�N� � DU �{�G�z��-0.02����{�G�z�����_Є0�.020� ��q@� � EU {�G�z��-0.02 {�G�z�� �Zd;�O��?0.184����Zd;�O��?�����0�.184� d�SF� � FU Zd;�O��?0.184 Zd;�O��? ������d�C�scores�for�sentence� U�D" " H � �-7��������UЈ̄7�.0� |�k=" " JU  �-7 � � �-7��������|Ј̄7�.0� \�K=" " LU  �-7 � ����\Ј�+34�.0� ]�L" " NU �4�-20����4����]Ј̄20�.0� |�k>" " PU 4�-20 4� � 0�������|Ј�0�.0�RHF" " RU  0  R����a����&��U��&��V�Words�in�text�are�looked�up�in�a�computerized�dictionary�and�their�idea�category�is�identified.��Probability�of�occurrence� � � S �of�that�category�in�each�"marker"�context�is�added�to�the�accumulating�context�sums.��In�the�illustration�above,�deviations�from�the�mean�probability�of�occurrence�for�that�category�are�summed.��Negative�figures�reflect�below-mean�deviations.��Positive�figures�indicate�above-mean�deviations�or�an�emphasis�on�that�contextual�approach.��Final�C-scores�for�the�entire�sentence�(or�text)�are�multiplied�by�a�constant,�and�their�mean�for�a�text�is�set�at�zero.TABLE B2/D<M�� << deUU �� �/1�#���&��V��&��������&��V��  ��  �� �GT����Table��  �2�  ���.���  �Accumulation�of�C-scores�for�an�Illustrative�Sentence��a��(n�$��� �(�  �1�  �)� �F/GWA�� << B���Level 1Level 2Level 3Level 4Level 5("T�$����Table��  �1�  ���.�� d� �81�"���&��V��&��������&��V��  ��  ���&��V������*s�IO ddd Xdd Xdd X��s��,+dd ��,�dd0���,�dd0���,�dd0���,�dd0���,�dd0���+  ��   � ��� 0�dd  ����0�Text�groups� 0&dd ������0�� SIUU � �?1�������?���S�1�� uBUUU  �?1 �? � @2������@����2�� uBUUU  @2 @ � @3������@����3�� uBUUU  @3 @ � @4������@����4�� uBUUU  @4 @ � @5������@����5�� b�UHUUU  @5 @ ������b�1.�Scholarly�article� 7�*  �������7Є��� f�Y# U� �������$@10.4�������������$@���f�10�.4� ��{E U� ������$@10.4 ������$@ �fffff&K@54.3�������fffff&K@�����54�.3� ��{E U� fffff&K@54.3 fffff&K@ �ffffff>@30.4�������ffffff>@�����30�.4� ��{E U� ffffff>@30.4 ffffff>@ �����̌F@45.1�����������̌F@�����45�.1� bXKU� ����̌F@45.1 ����̌F@ ������b�2.�Airline�finance�news� VL�� �������$@10.4�������������$@���V�10�.4� cYE��U� ������$@10.4 ������$@ �������cЄ��� cY#��U� �@G@46.5�������@G@���c�46�.5� �{E��U� @G@46.5 @G@ �������9@25.6�������������9@�����25�.6� �{E��U� ������9@25.6 ������9@ �333333C@38.4�������333333C@�����38�.4� bXK��U� 333333C@38.4 333333C@ ������b�3.�Magazine�fiction� VL�� �fffff&K@54.3�������fffff&K@���V�54�.3� �{E��U� fffff&K@54.3 fffff&K@ �@G@46.5�������@G@�����46�.5� cYE��U� @G@46.5 @G@ �������cЄ��� aW#��U� � <@28�������<@���a�28�.0� �yC��U�  <@28 <@ �������(@12.3�������������(@�����12�.3� bXK��U� ������(@12.3 ������(@ ������b�4.�Religious�devotional� VLYY �ffffff>@30.4�������ffffff>@���V�30�.4� �{EYYU� ffffff>@30.4 ffffff>@ �������9@25.6�������������9@�����25�.6� �yEYYU� ������9@25.6 ������9@ � <@28�������<@�����28�.0� aWCYYU�  <@28 <@ �������aЄ��� cY#YYU� �ffffff0@16.4�������ffffff0@���c�16�.4� bXKYY U� ffffff0@16.4 ffffff0@ ����� �b�5.�Crime�editorial� VL!  �����̌F@45.1�����������̌F@�� �V�45�.1� �{E"U� ����̌F@45.1 ����̌F@  �333333C@38.4�������333333C@�� ���38�.4� �{E#U� 333333C@38.4 333333C@  �������(@12.3�������������(@�� ���12�.3� �{E$U� ������(@12.3 ������(@  �ffffff0@16.4�������ffffff0@�� ���16�.4� cYE%U� ffffff0@16.4 ffffff0@  �������cЄ���9/-&U�   9� �� �81�"���&��V��&��������&��V��  ��  ���&��V������*s�KL ddd Xdd Xdd X��s��, dd ��,�dd0��,�dd0��,�dd0��,�dd0��+  ��   � ��� 0�dd  ����0�Four�context�dimensions� 0&dd ������0�� .$UU ����.�Traditional� 5+UU3 ����5�Practical� 5+UU3 ����5�Emotional� 5+UU3 ����5�Analytic� =�0#UU3  ������=�1.�Scholarly�article� T�G ��Q��� �-3.24�����Q��� ����TЄ3�.24� ~�q@ U �Q��� �-3.24 �Q��� � � ףp= $@10.02���� ףp= $@���~�10�.02� �r@ U  ףp= $@10.02  ףp= $@ ��Q���,�-14.46�����Q���,����Є14�.46� ~�qA U �Q���,�-14.46 �Q���,� ��G�z�@7.67�����G�z�@���~�7�.67� \RE U �G�z�@7.67 �G�z�@ ������\�2.�Airline�finance�news� QG��  �\���(\�-5.59����\���(\����QЄ5�.59� {q@��U \���(\�-5.59 \���(\� ��Q��k&@11.21�����Q��k&@���{�11�.21� {q@��U �Q��k&@11.21 �Q��k&@ ��G�z�-6.77�����G�z����{Є6�.77� zp@��U �G�z�-6.77 �G�z� �ffffff�?1.15����ffffff�?���z�1�.15� \RE��U ffffff�?1.15 ffffff�? ������\�3.�Magazine�fiction� QG�� ���Q���-2.69������Q������QЄ2�.69� |r@��U ��Q���-2.69 ��Q��� ��G�z�+�-13.79�����G�z�+����|Є13�.79� |rA��U �G�z�+�-13.79 �G�z�+� �)\����=@29.66����)\����=@���|�29�.66� |r@��U )\����=@29.66 )\����=@ ��z�Ga*�-13.19�����z�Ga*����|Є13�.19� ^TG��U �z�Ga*�-13.19 �z�Ga*� ������^�4.�Religious�devotional� PFYY �R���Q@6.08����R���Q@���P�6�.08� zp?YYU R���Q@6.08 R���Q@ ���(\���-7.19������(\������zЄ7�.19� zp@YYU ��(\���-7.19 ��(\��� ���(\��@5.44������(\��@���z�5�.44� zp?YYU ��(\��@5.44 ��(\��@ �R���Q�-4.33����R���Q����zЄ4�.33� ]SFYYU R���Q�-4.33 R���Q� ����� �]�5.�Editorial�on�recent�times� PF  ��G�z @3.51�����G�z @�� �P�3�.51� {q?U �G�z @3.51 �G�z @  ������L)�-12.65���������L)��� �{Є12�.65� |rAU �����L)�-12.65 �����L)�  ����Qx3@19.47�������Qx3@�� �|�19�.47� |r@U ���Qx3@19.47 ���Qx3@  �)\����$�-10.33����)\����$����|Є10�.33�WMK U )\����$�-10.33  )\����$� W�TABLE B +� �/1�#���&��V��&��������&��V��  ��  �� �GT����Table��  �3�  ���.���  ���&��V�����Distribution�of�contextual�scores�(C-scores)�for�five�selected�texts�#�����&��V�#�I%I�I�?I@I@I@I@I�?I�I������$@�Ifffff&K@�Iffffff>@�I����̌F@�I@I������$@�I�I@G@�I������9@�I333333C@�I@Ifffff&K@�I@G@�I�I<@�I������(@�I@Iffffff>@�I������9@�I<@�I�Iffffff0@�I@I����̌F@�I333333C@�I������(@�Iffffff0@�I��TABLE BITABLE B � �/1�#���&��V��&��������&��V��  ��  �� �GT����Table��  �4�  ���.���  �Matrix�of�Euclidean�distances�between�five�texts�on�the�four�contextual�  �dimensionsHC%�GFd:~� �,�'��D� �V� �81�"���&��V��&��������&��V��  ��  ���&��V������*s�ST ddd Xdd Xdd X��s��,+dd ��,�dd0���,�dd0���,�dd0���,�dd0���,�dd0���+  ��   �����Conceptual�Categories��b��� A�0'%%   ����A�Five�selected�texts��a��� 4*dd ���������4�� VLUU� � �?1�������?���V�1�� uBUUU  �?1 �? � @2������@����2�� uBUUU  @2 @ � @3������@����3�� uBUUU  @3 @ � @4������@����4�� uBUUU  @4 @ � @5������@����5�� b�UHUUU  @5 @ ������b�Activity� Y�L``  �������'@11.8�������������'@���Y�11�.80� ��|E`` U� ������'@11.8 ������'@ �H�z�Gѿ-0.27�������H�z�Gѿ�����-0�.27� ��}F`` U� H�z�Gѿ-0.27 H�z�Gѿ � ףp= ׿-0.36������� ףp= ׿�����-0�.36� ��}F`` U�  ףp= ׿-0.36  ףp= ׿ ���(\�� �-3.47���������(\�� ������-3�.47� ��}F`` U� ��(\�� �-3.47 ��(\�� � ��p= ף��-1.54��������p= ף�������-1�.54� cYL``U� �p= ף��-1.54 �p= ף�� ������c�Fellow�feeling� VL!! ��G�z�#@9.79��������G�z�#@���V�9�.79� �|E!!U� �G�z�#@9.79 �G�z�#@ ��G�z�-2.76��������G�z������-2�.76� �}F!!U� �G�z�-2.76 �G�z� ��G�z�-2.76��������G�z������-2�.76� �}F!!U� �G�z�-2.76 �G�z� ��G�z�-2.76��������G�z������-2�.76� �|F!!U� �G�z�-2.76 �G�z� ����Q�@7.43����������Q�@�����7�.43� bXK!!U� ���Q�@7.43 ���Q�@ ���<���b�Kind�of� VL�� ��z�G�@5.47��������z�G�@���V�5�.47� �|E��U� �z�G�@5.47 �z�G�@ ��(\����-2.87��������(\���������-2�.87� �|F��U� �(\����-2.87 �(\���� ���Q��?1.57���������Q��?�����1�.57� �|E��U� ��Q��?1.57 ��Q��? ��Q����-4.98��������Q���������-4�.98� �}F��U� �Q����-4.98 �Q���� ���Q���-1.57���������Q��������-1�.57� cYL��U� ��Q���-1.57 < ��Q��� ������c�Now� VL�� ���(\�� @3.72���������(\�� @���V�3�.72� �zE��U� ��(\�� @3.72 ��(\�� @ �333333"@9.1�������333333"@�����9�.10� �{D��U� 333333"@9.1 333333"@ � ףp= ǿ-0.18������� ףp= ǿ�����-0�.18� �}F��U�  ףp= ǿ-0.18  ףp= ǿ �R���Q�-2.79�������R���Q������-2�.79� �}F��U� R���Q�-2.79 R���Q� ��Q�����-1.87��������Q����������-1�.87� cYL�� U� �Q�����-1.87 �Q����� ����� �c�Object� WM��!  �{�G�z�-0.64�������{�G�z��� �W�-0�.64� �|F��"U� {�G�z�-0.64 {�G�z�  �{�G�z"@9.24�������{�G�z"@�� ���9�.24� �|E��#U� {�G�z"@9.24 {�G�z"@  ���Q�տ-0.33���������Q�տ�� ���-0�.33� �|F��$U� ��Q�տ-0.33 ��Q�տ  �)\���(�?0.11�������)\���(�?�� ���0�.11� �|E��%U� )\���(�?0.11 )\���(�?  ��p= ף��-1.79��������p= ף�������-1�.79� cYL��&U� �p= ף��-1.79 �p= ף�� ����� �c�Obligation� VLaa'  � @3.25������� @�� �V�3�.25� �{Eaa(U�  @3.25  @  �ףp= �"@9.42�������ףp= �"@�� ���9�.42� �|Eaa)U� ףp= �"@9.42 ףp= �"@  �ףp= ���-1.49�������ףp= ����� ���-1�.49� �}Faa*U� ףp= ���-1.49 ףp= ���  �ףp= ���-1.49�������ףp= ����� ���-1�.49� �}Faa+U� ףp= ���-1.49 ףp= ���  �ףp= ���-1.49�������ףp= ��������-1�.49� cYLaa,U� ףp= ���-1.49 ףp= ��� ���<�� �c�Spatial� WM""-  ���-1.25����������� �W�-1�.25� �}F"".U� ��-1.25 ��  �)\����%@10.83�������)\����%@�� ���10�.83� �|F""/U� )\����%@10.83 )\����%@  �ffffff�?0.95�������ffffff�?�� ���0�.95� �|E""0U� ffffff�?0.95 ffffff�?  �H�z�G��-1.83�������H�z�G���� ���-1�.83� �}F""1U� H�z�G��-1.83 H�z�G��  � ףp= ��-1.44������� ףp= �������-1�.44� cYL""2U�  ףp= ��-1.44 <  ףp= �� ����� �c�About�changing� VL3  �fffff�!@8.95�������fffff�!@�� �V�8�.95� �|E4U� fffff�!@8.95 fffff�!@  ���(\���-4.44���������(\����� ���-4�.44� �|F5U� ��(\���-4.44 ��(\���  ��p= �##@9.57��������p= �##@�� ���9�.57� �|E6U� �p= �##@9.57 �p= �##@  ����(\�ҿ-0.29����������(\�ҿ�� ���-0�.29� �}F7U� ���(\�ҿ-0.29 ���(\�ҿ  �\���(\��-1.96�������\���(\�������-1�.96� cYL8U� \���(\��-1.96 \���(\�� ���<�� �c�Cognition� WM��9  ����Q� �-3.59����������Q� ��� �W�-3�.59� �}F��:U� ���Q� �-3.59 ���Q� �  ��������-2.95���������������� ���-2�.95� �}F��;U� �������-2.95 �������  ��z�Ga)@12.69��������z�Ga)@�� ���12�.69� �yF��<U� �z�Ga)@12.69 �z�Ga)@  � @2�������@�� ���2�.00� �xB��=U�  @2 @  ���Q��@5.88���������Q��@�����5�.88� bXK��>U� ��Q��@5.88 < ��Q��@ ����� �b�Eternal� WM� � ?  ��-4.25���������� �W�-4�.25� �}F� � @U� �-4.25 �  ���(\��տ-0.34���������(\��տ�� ���-0�.34� �{F� � AU� ��(\��տ-0.34 ��(\��տ  �������@5.9�������������@�� ���5�.90� �zD� � BU� ������@5.9 ������@  ���Q�@6.28���������Q�@�� ���6�.28� �|E� � CU� ��Q�@6.28 ��Q�@  �q= ףp�-2.68�������q= ףp������-2�.68� cYL� � DU� q= ףp�-2.68 q= ףp� ����� �c�Expression�arena� WM� � E  �333333 �-3.65�������333333 ��� �W�-3�.65� �|F� � FU� 333333 �-3.65 333333 �  �H�z�G�?1.33�������H�z�G�?�� ���1�.33� �|E� � GU� H�z�G�?1.33 H�z�G�?  �R���Q�-2.04�������R���Q��� ���-2�.04� �{F� � HU� R���Q�-2.04 R���Q�  �������@6.4�������������@�� ���6�.40� �{D� � IU� ������@6.4 ������@  �333333�-2.65�������333333������-2�.65� cYL� � JU� 333333�-2.65 333333� ����� �c�We� WM_ _ K  ���(\���-2.47���������(\����� �W�-2�.47� �}F_ _ LU� ��(\���-2.47 ��(\���  ���������-1.85����������������� ���-1�.85� �}F_ _ MU� ��������-1.85 ��������  ����Q���-1.67����������Q����� ���-1�.67� �|F_ _ NU� ���Q���-1.67 ���Q���  ���(\�� @8.38���������(\�� @�� ���8�.38� �|E_ _ OU� ��(\�� @8.38 ��(\�� @  ���(\���-2.47���������(\��������-2�.47� cYL_ _ PU� ��(\���-2.47 ��(\��� ����� �c�Want�responding� WM Q  �)\���(�-2.27�������)\���(��� �W�-2�.27� �}F RU� )\���(�-2.27 )\���(�  ��G�z� �-3.21��������G�z� ��� ���-3�.21� �|F SU� �G�z� �-3.21 �G�z� �  �������ٿ-0.4�������������ٿ�� ���-0�.40� �{E TU� ������ٿ-0.4 ������ٿ  �R����#@9.91�������R����#@�� ���9�.91� �|E UU� R����#@9.91 R����#@  �\���(\��-1.46�������\���(\�������-1�.46� cYL VU� \���(\��-1.46 \���(\�� ����� �c�Depressed� WM� � W  �= ףp=�-2.28�������= ףp=��� �W�-2�.28� �}F� � XU� = ףp=�-2.28 = ףp=�  �= ףp=�-2.28�������= ףp=��� ���-2�.28� �}F� � YU� = ףp=�-2.28 = ףp=�  �= ףp=�-2.28�������= ףp=��� ���-2�.28� �}F� � ZU� = ףp=�-2.28 = ףp=�  ��Q���$@10.46��������Q���$@�� ���10�.46� �}F� � [U� �Q���$@10.46 �Q���$@  �= ףp=�-2.28�������= ףp=������-2�.28� cYL� � \U� = ףp=�-2.28 = ףp=� ����� �c�With� VL� � ]  �)\���(�?0.44�������)\���(�?�� �V�0�.44� �{E� � ^U� )\���(�?0.44 )\���(�?  ��p= ף@5.16��������p= ף@�� ���5�.16� �|E� � _U� �p= ף@5.16 �p= ף@  �{�G�z��-0.01�������{�G�z���� ���-0�.01� �}F� � `U� {�G�z��-0.01 {�G�z��  ��G�z&@11.04��������G�z&@�� ���11�.04� �|F� � aU� �G�z&@11.04 �G�z&@  �ףp= �@2.98�������ףp= �@�����2�.98� bXK� � bU� ףp= �@2.98 ףp= �@ ����� �b�Good� WMccc  �\���(\�-6.09�������\���(\��� �W�-6�.09� �|FccdU� \���(\�-6.09 \���(\�  �ffffff�?0.95�������ffffff�?�� ���0�.95� �{EcceU� ffffff�?0.95 ffffff�?  �q= ףp@2.18�������q= ףp@�� ���2�.18� �|EccfU� q= ףp@2.18 q= ףp@  �R���Q*@13.16�������R���Q*@�� ���13�.16� �}FccgU� R���Q*@13.16 R���Q*@  ��G�z �-3.51��������G�z ������-3�.51� cYLcchU� �G�z �-3.51 �G�z � ����� �c�Begin�action� VL$$i  ���(\���?1.86���������(\���?�� �V�1�.86� �{E$$jU� ��(\���?1.86 ��(\���?  �R���Q@2.79�������R���Q@�� ���2�.79� �{E$$kU� R���Q@2.79 R���Q@  ��z�G�@2.61��������z�G�@�� ���2�.61� �{E$$lU� �z�G�@2.61 �z�G�@  �������-@14.9�������������-@�� ���14�.90� �|E$$mU� ������-@14.9 ������-@  �{�G�z��-0.04�������{�G�z�������-0�.04� cYL$$nU� {�G�z��-0.04 {�G�z�� ���<�� �c�Error� VL��o  ��z�G�@4.22��������z�G�@�� �V�4�.22� �|E��pU� �z�G�@4.22 �z�G�@  ��G�z���-1.98��������G�z����� ���-1�.98� �}F��qU� �G�z���-1.98 �G�z���  ��G�z���-1.98��������G�z����� ���-1�.98� �}F��rU� �G�z���-1.98 �G�z���  �@1@17.25�������@1@�� ���17�.25� �|F��sU� @1@17.25 @1@  �{�G�z�?0.89�������{�G�z�?�����0�.89� bXK��tU� {�G�z�?0.89 < {�G�z�? ����� �b�Tender� WM��u  ��(\����-3.12��������(\������ �W�-3�.12� �}F��vU� �(\����-3.12 �(\����  ��(\����-3.12��������(\������ ���-3�.12� �}F��wU� �(\����-3.12 �(\����  ��(\����-0.78��������(\������ ���-0�.78� �}F��xU� �(\����-0.78 �(\����  ��(\����-3.12��������(\������ ���-3�.12� �|F��yU� �(\����-3.12 �(\����  ���Q�@7.03���������Q�@�����7�.03� bXK��zU� ��Q�@7.03 ��Q�@ ����� �b�Tradition�directed� VL��{  ��Q���@5.98��������Q���@�� �V�5�.98� �{E��|U� �Q���@5.98 �Q���@  �333333�-3.9�������333333��� ���-3�.90� �|E��}U� 333333�-3.9 333333�  � ףp= ��-1.44������� ףp= ���� ���-1�.44� �|F��~U�  ףp= ��-1.44  ףp= ��  �\���(\�?1.21�������\���(\�?�� ���1�.21� �{E��U� \���(\�?1.21 \���(\�?  ��p= ף @8.32��������p= ף @�����8�.32� bXK���U� �p= ף @8.32 �p= ף @ ����� �b�Bad� WMdd�  �q= ףp �-3.18�������q= ףp ��� �W�-3�.18� �}Fdd�U� q= ףp �-3.18 q= ףp �  �q= ףp �-3.18�������q= ףp ��� ���-3�.18� �|Fdd�U� q= ףp �-3.18 q= ףp �  ���Q�@4.53���������Q�@�� ���4�.53� �|Edd�U� ��Q�@4.53 ��Q�@  �q= ףp �-3.18�������q= ףp ��� ���-3�.18� �|Fdd�U� q= ףp �-3.18 q= ףp �  ����(\�!@8.78����������(\�!@�����8�.78� bXKdd�U� ���(\�!@8.78 ���(\�!@ ����� �b�Human�roles� WM%%�  ����(\��-0.58����������(\���� �W�-0�.58� �}F%%�U� ���(\��-0.58 ���(\��  ��(\��� �-3.37��������(\��� ��� ���-3�.37� �}F%%�U� �(\��� �-3.37 �(\��� �  ��(\�����-1.81��������(\������� ���-1�.81� �}F%%�U� �(\�����-1.81 �(\�����  �R���Q�-2.29�������R���Q��� ���-2�.29� �|F%%�U� R���Q�-2.29 R���Q�  ��G�z.!@8.59��������G�z.!@�����8�.59� bXK%%�U� �G�z.!@8.59 �G�z.!@ ����� �b�Status� VL���  ������� �-3.2������������� ��� �V�-3�.20� �{E���U� ������ �-3.2 ������ �  ������� �-3.2������������� ��� ���-3�.20� �{E���U� ������ �-3.2 ������ �  ������� �-3.2������������� ��� ���-3�.20� �{E���U� ������ �-3.2 ������ �  ������� �-3.2������������� ��� ���-3�.20� �|E���U� ������ �-3.2 ������ �  ���Q�^0@16.37���������Q�^0@�����16�.37� cYL���U� ��Q�^0@16.37 ��Q�^0@ ����� �c�Sin� WM���  ��G�z�-2.26��������G�z��� �W�-2�.26� �}F���U� �G�z�-2.26 �G�z�  ��G�z�-2.26��������G�z��� ���-2�.26� �}F���U� �G�z�-2.26 �G�z�  ��G�z�-2.26��������G�z��� ���-2�.26� �}F���U� �G�z�-2.26 �G�z�  ��G�z�-2.26��������G�z��� ���-2�.26� �}F���U� �G�z�-2.26 �G�z�  ��Q���0@16.92��������Q���0@�����16�.92� cYL���U� �Q���0@16.92 �Q���0@ ���<�� �c�Sanction� WMhh�  ����Q��-5.68����������Q���� �W�-5�.68� �}Fhh�U� ���Q��-5.68 ���Q��  ����Q��-5.68����������Q���� ���-5�.68� �|Fhh�U� ���Q��-5.68 ���Q��  �q= ףp@4.11�������q= ףp@�� ���4�.11� �|Ehh�U� q= ףp@4.11 q= ףp@  ����Q��-5.68����������Q���� ���-5�.68� �}Fhh�U� ���Q��-5.68 ���Q��  ��z�G�H@49.01��������z�G�H@�����49�.01� cYLhh�U� �z�G�H@49.01 < �z�G�H@ ����� �c�Total�words�in�text� UKee�  ��@541�������@�� �U�541�� �yDee�U �@541 �@  �`}@470������`}@�� ���470�� �yDee�U `}@470 `}@  ���@724��������@�� ���724�� �yDee�U ��@724 ��@  ��u@349�������u@�� ���349�� �zDee�U �u@349 �u@  �<�@1167������<�@�����1167�� i�XKee�U <�@1167 <�@ ������i�Percent�unclassified� \�K&&� �.@15�������333333�?���\�15%�� ��yD&&�U� .@15 333333�? �9@25��������?�����25%�� ��yD&&�U� 9@25 �? �0@16�������{�G�z�?�����16%�� ��xD&&�U� 0@16 {�G�z�? �  @8�������{�G�z�?�����8%�� ��xC&&�U�   @8 {�G�z�? �.@15�������333333�?�����15%��ZPN&&�U� .@15  333333�? Z���a��Texts�are�those�referred�to�in�Tables�3�and�4:��(1)�Scholarly�article;�(2)�Airline�news;�(3)�Magazine�fiction;�(4)�Religious� ��� �devotional;�(5)�Crime�editorial.���b��Heuristic�category�labels�suggest�the�overall�character�of�word�meanings�in�a�category.��Examination�of�words�and� ii� �surrounding�phrases�refines�the�identity�of�the�subset�of�words�which�were�actually�found�in�the�text.TABLE B �� �/1�#���&��V��&��������&��V��  ��  �� �GT����Table��  �5�  ���.���  �E-Scores�for�five�texts�on�selected�conceptual�categoriesKKKKKKK�?K�Q��� �K ףp= $@K�Q���,�K�G�z�@K@K\���(\�K�Q��k&@K�G�z�Kffffff�?K@K��Q���K�G�z�+�K)\����=@K�z�Ga*�K@KR���Q@K��(\���K��(\��@KR���Q�K@K�G�z @K�����L)�K���Qx3@K)\����$��TABLE BKE"Thin Top/Bottom dxdS�SS�S�?S@S@S@S@SS������'@�SH�z�Gѿ�S ףp= ׿�S��(\�� ��S�p= ף���SS�G�z�#@�S�G�z��S�G�z��S�G�z��S���Q�@�SS�z�G�@�S�(\�����S��Q��?�S�Q�����S��Q����SS��(\�� @�S333333"@�S ףp= ǿ�SR���Q��S�Q������SS{�G�z��S{�G�z"@�S��Q�տ�S)\���(�?�S�p= ף���SS @�Sףp= �"@�Sףp= ����Sףp= ����Sףp= ����SS���S)\����%@�Sffffff�?�SH�z�G���S ףp= ���S S fffff�!@�S ��(\����S �p= �##@�S ���(\�ҿ�S \���(\���S S ���Q� ��S ��������S �z�Ga)@�S @�S ��Q��@�S S ��S ��(\��տ�S ������@�S ��Q�@�S q= ףp��S S 333333 ��S H�z�G�?�S R���Q��S ������@�S 333333��S S ��(\����S ���������S ���Q����S ��(\�� @�S ��(\����SS)\���(��S�G�z� ��S������ٿ�SR����#@�S\���(\���SS= ףp=��S= ףp=��S= ףp=��S�Q���$@�S= ףp=��SS)\���(�?�S�p= ף@�S{�G�z���S�G�z&@�Sףp= �@�SS\���(\��Sffffff�?�Sq= ףp@�SR���Q*@�S�G�z ��SS��(\���?�SR���Q@�S�z�G�@�S������-@�S{�G�z���SS�z�G�@�S�G�z����S�G�z����S@1@�S{�G�z�?�SS�(\�����S�(\�����S�(\�����S�(\�����S��Q�@�SS�Q���@�S333333��S ףp= ���S\���(\�?�S�p= ף @�SSq= ףp ��Sq= ףp ��S��Q�@�Sq= ףp ��S���(\�!@�SS���(\���S�(\��� ��S�(\������SR���Q��S�G�z.!@�SS������ ��S������ ��S������ ��S������ ��S��Q�^0@�SS�G�z��S�G�z��S�G�z��S�G�z��S�Q���0@�SS���Q���S���Q���Sq= ףp@�S���Q���S�z�G�H@�SS�@S`}@S��@S�u@S<�@SS/333333�?.@15�S/�?9@25�S/{�G�z�?0@16�S.{�G�z�?  @8�S/333333�?.@15��TABLE BS� �%��!���&��V��&��������&��V��  ��  ���'��V�������'X��X�'��V��  ��  ���)�����!��������L���B.��@�� H ��P�X� `"��XB������'��U�X'X��������Quality�&�Quantity��� �24� �:�245�265,�1990.� � ����1990���Kluwer�Academic�Publishers.��Printed�in�the�Netherlands.��� u� �����#�'X��X�'��UI#�������Method�of�inquiry�paper��� o ��� �Contextual�Content�Analysis� ���1��� �? ���DONALD�G.�MCTAVISH*�&�ELLEN�B.�PIRRO**�*��Department�of�Sociology,�University�of�Minnesota,�Minneapolis,�MN,�USA��� � �  �**��Department�of�Political�Science,�Iowa�State�Univerisity,�Ames,�IA,�USA��� } �  ���B.��@�� H ��P�X� `"���XB���'��U�X'X���� �Abstract.� �This�article�suggests�one�way�to�systematically�code�textual�data�for�research.��The�approach�utilizes�computer� O�  �content�analysis�to�examine�patterns�of�emphasized�ideas�in�text�as�well�as�the�social�context�or�underlying�perspective�reflected�in�the�text.��A�conceptual�dictionary�is�used�to�organize�word�meanings.��An�extensive�profile�of�word�meanings�is�used�to�characterize�and�discriminate�social�contexts.��Social�contexts�are�analyzed�in�relation�to�four�reference�dimensions�(traditional,�practical,�emotional�and�analytic)�which�are�used�in�the�social�science�literature.��The�approach�is�illustrated�with�five�widely�varying�texts,�analyzed�with�selected�comparative�data.��This�approach�has�been�useful�in�many�social�science�investigations�to�systematically�score�open-ended�textual�information.��Scores�can�be�incorporated�into�quantitative�analysis�with�other�data,�used�as�a�guide�to�qualitative�studies,�and�to�help�integrate�strengths�of�quantitative�and�qualitative�approaches�to�research.���� �Abbreviations:� �C�scores�=�Contextual�Scores,�E�scores�=�Idea/Emphasis�Scores,�KWIC�=�Key�Word�In�Context,�MCCA� �- �=�Minnesota�Contextual�Content�Analysis��#�'X��X�'��U�#���B.��@�� H ��P�X� `"���XB�� �Introduction� �� L� �������Words�are�a�basic�form�of�data�for�much�social�science�research�because�they�are�the�usual�medium�for�social�exchange.��For�many�purposes,�insight�into�meanings�can�be�obtained�by�examining�profiles�of�ideas�and�contextual�information�contained�in�text.��In�this�paper�we�address�one�approach�to�the�social�science�research�problem�of�systematically�coding�textual�data.��The�approach�makes�quantitative�distinctions�between�texts�varying�in�both�the�pattern�of�emphasis�upon�different�sets�of�ideas�and�in�the�context�or�social�perspective�from�which�these�ideas�are�addressed.��Scores�are�used�to�describe�comparative�patterns�of�meaning�in�textual�data,�generate�traditional�statistical�analyses�with�other�non����񼼼���textual�variables,�and�aid�in�organizing�and�focusing�further�qualitative�analysis.�� @ �By�"text"�we�mean�a�transcript�of�naturally�occurring�verbal�material.��Included�are�conversations,�written�documents�such�as�diaries�or�organization�reports,�books,�written�or�taped�responses�to�open-ended�questions,�media�recordings,�and�verbal�descriptions�of�observations.��Ultimately,�the�transcript�consists�of�a�computerized�file�of�conventional�words�and�sentences�for�one�or�more�cases.�� @ �Methodologies�for�directly,�systematically�and�efficiently�handling�textual�data�are�needed.��Traditionally,�trained�coders�are�utilized�but�serious�validity,�reliability,�and�practical�problems�are�often�encountered.��Computer�approaches�(available�since�the�1960's)�permit�more�systematic�and�reliable�coding�of�themes�and�meanings�in�text�but�these�have�not�been�widely�adopted�in�social�science�research.�Like�Markoff�et�al.�(1975),�we�take�the�view�that�content�analysis�must�be�integrated�with� �*�%,  �traditional�methodology.��The�approach�described�here�extends�computer�content�analysis,�making�it�a�more�useful�and�complementary�research�tool�in�traditional�social�science�methodology.�� @ �Content�analysis�has�been�summarized�in�a�number�of�places�in�the�literature.��2����Contextual��� #� �content�analysis,�implemented�in�the�Minnesota�Contextual�Content�Analysis�(MCCA)�computer�program,�builds�upon�computer�content�analysis�methodology�in�a�number�of�ways.��3��� � � ���� �Overview�of�Contextual�Content�Analysis� �� � _ ��First,���all��words�in�one�or�more�texts�are�divided�into�a�large�number�of�idea�categories�(including�a�"not� �2  �elsewhere�classified"�category),�guided�by�a�conceptual�"dictionary".�A�dictionary�groups�words�(or�in�our�case,�word�meanings)�into�categories�thought�to�express�(singly�or�in�patterns)�ideas�important�to�an�investigator.�Several�conceptual�dictionaries�have�been�used�in�computer-based�content�analysis,�each�organized�around�somewhat�different�theoretical�perspectives.��4��� '�  �� @ �The�contextual-conceptual�dictionary�MCCA�uses�is�oriented�toward�more�frequently�used�words�whose�meanings�are�organized�into�a�large�number�of�categories.��The�categories�are�of�general�social�science�interest�and�are�mutually�exclusive�(Pirro�and�McTavish�1982).��Words�with�multiple�meanings�are�disambiguated.��Relative�emphasis�upon�each�category�in�the�text�is�then�normed�with�respect�to�a�standard�(i.e.,�the�expected�emphasis�on�these�categories,�accounting�for�expected�variability�in�the�use�of�a�category�over�a�number�of�social�contexts),�a�process�described�later.��This�vector�of�normed�scores�(called�"emphasis"�scores�or�E-scores)�permits�an�investigator�to�examine�the�over��and�under�emphasis�on�idea�categories�relative�to�the�norm�of�expected�category�usage.��Broader�concepts�and�themes�in�a�text�can�be�identified�from�scores�for�sets�of�related�categories.��Quantitative�distinctions�between�texts�can�also�be�made�in�terms�of�the�overall�profile�of�emphasis�on�idea�categories.������ @ �Secondly,�MCCA�incorporates�an�hypothesis�that�different�social�contexts�(social�groups,�institutions,�organizational�cultures,�or�other�socially�defined�situations)�can�be�identified�by�the�overall�profile�of�relative�emphasis�upon�idea�categories�utilized�in�communication�in�that�context.��The�idea-emphasis�profile�appears�to�contain�important�information�for�distinguishing�and�characterizing�social�contexts.�� @ �To�aid�in�interpreting�contextual�information�in�these�profiles,�we�use�four�general�"marker"�contexts�we�call�"traditional",�"practical",�"emotional",�and�"analytic".��Each�marker�context�is�an�experimental,�empirically-derived�profile�of�relative�emphasis�on�each�idea�category,�which�characterizes�the�perspective�typical�of�a�general�social�or�institutional�context.��As�a�set,�the�four�contextual�markers�serve�as�dimensions�to�define�a�social�context�space.��MCCA�computes�these�contextual�scores�(called�C�scores).��Texts�can�be�scored�and�differentiated�on�these�four�dimensions.��Distinctions�can�be�made,�for�example,�between�a�more�"traditional"�concern�for�breach�of�norms�and�appropriate�sanctions�in�a�religious�discussion,�and�a�more�"practical"�concern�for�failure�to�successfully�achieve�goals�and�consequences�in�a�business�discussion.�Similar�ideas�may�be�discussed�in�quite�different�ways�in�different�social�contexts.��These�scores�appear�to�be�important�parameters�of�social�contexts.�� @ �Third,�the�approach�helps�link�strengths�of�qualitative�and�quantitative�social�science�research.��For�example,�an�investigator�can�realistically�examine�transcribed�conversational�interviews�on�a�topic�for�a�large�representative�sample�of�cases.��Quantitative�scores�can�help�guide�comparative,�qualitative�analysis�of�social�meanings�in�textual�data,�adding�depth�and�anchoring�to�quantitative�causal�analyses� u-$'+ �as�well.��Fourth,�computerized�content�analysis�eliminates�coder�reliability�problems,�permitting�more�careful�analysis�of�measurement�and�validity�issues.�� @ �In�summary,�this�approach�has�a�number�of�advantages�for�systematic�analysis.��Norming�provides�a�basis�for�examining�topical�emphasis�(including�distinctive�omissions)�in�a�text,�a�task�which�is�problematic�in�coding�open-ended�response�data�with�earlier�procedures.�Using�the�normed,�idea-emphasis�scores�(E-scores)�and�scores�reflecting�emphasis�upon�the�set�of�four�marker�contexts�(C-scores),�naturally�occurring�textual�material�can�be�"coded"�to�reflect�meanings�of�interest�to�an�investigator.�The�set�of�scores�(C-scores�and�E-scores�for�each�text)�can�be�combined�for�traditional�quantitative,�statistical�analysis�with�independent�and�dependent�variables�measured�in�other�ways.�We�refer�to�this�as�"Contextual�Content�Analysis"�to�distinguish�it�from�more�traditional�hand�and�computer�content�analytic�approaches.��This�approach�has�utility�for�social�science�research�by�providing�a�broad�framework�for�characterizing�social�meanings�in�text�and�a�practical,�systematic�means�for�scoring�textual�data.�The�following�sections�further�describe�MCCA�and�provide�illustrations�of�its�use.���� �Meanings�in�Text� �� ��  ��As�in�any�research,�the�meaning�that�is�attributed�to�a�text�depends�upon�the�researcher's�theory.��There�is�no�general�answer�to�the�question�of�what�a�textual�passage�"really"�means.��Nor�is�there�generally�a�research�interest�in�capturing�"all"�of�the�meanings�that�may�be�attributable�to�a�text.�In�short,�the�research�problem�and�the�theory�the�investigator�uses�will�specify�the�relevant�meanings�in�appropriate�text�for�certain�research�purposes.��� @ �Markoff�et�al.�(1975)�distinguishes�between�a�situation�where�subjects�have�an�interest�in�sharing�meanings�and�a�situation�where�subjects�intend�to�manipulate�the�investigator's�understanding�by�what�is�said.��From�our�point�of�view,�manipulative�intentions�on�the�part�of�subjects�do�not�invalidate�an�analysis�of�what�is�said.��This�does�suggest,�however,�that�explanatory�theories�might�also�include�the�possibility�of�intentional�manipulation.��Similarly,�sub-cultural�and�individualistic�uses�of�words�should�also�be�entertained�in�explanatory�research�uses�of�text.���� �Measuring�Context� �� �!O ��By�"context",�we�mean�the�shared�meaning�or�social�definition�of�a�situation�of�interaction.�Context�provides�an�underlying�orientation�for�subsequent�action.��5��There�are�several�levels�of�context.��Broad� \$ ! �social�contexts�may�be�all-encompassing�such�as�the�meaning�of�being�a�world�citizen�or�a�human�or�a�member�of�a�culture�or�nation�or�sub-culture.��More�specifically,�shared�contexts�exist�regarding�aspects�of�life,�such�as�work,�family�or�leisure.��6��The�meaning�of�social�context�plays�an�important�role� '� $ �in�a�number�of�explanatory�perspectives�in�social�science.��7��� (�!% �� @ �The�setting�of�communication�provides�a�framework�within�which�other�types�of�analysis�proceed.��Typically,�in�hand�content�analysis,�context�information�is�assumed�or�intuitively�determined�(e.g.�"since�we�are�in�a�work�context,�we�will�examine�meanings�of�job�satisfaction�and�not�religious�satisfaction"),�or�uses�information�outside�the�communication�itself�(e.g.�the�status�of�the�speaker,�conditions�under�which�the�text�was�prepared)�(Krippendorff�1980).��This�confounds�description�or�characterization�of�a�communication�with�the�explanatory�problem�of�determining�its�causes�and� v-%'+ �consequences.��To�avoid�these�hazards,�one�could�focus�on�the�measurement�problem:�coding�descriptive�information�about�ideas�and�context�expressed�within�a�text,�then�utilize�some�of�these�codes�in�explanatory�analysis�with�independently�measured�variables.���� @ �Typically,�words�introduce�the�context,�although�other�signs�and�symbols�may�do�so�as�well.��For�example,�when�someone�says�"Tell�me�about�your�work",�the�conversation�has�had�limits�and�direction�established�by�placement�within�one�context�(an�economic�or�work�context)�to�the�relative�exclusion�of�other�contexts�(e.g.�religious,�family,�leisure).��� @ �Context�is�indicated�by�several�features�of�language.��First,�it�is�indicated�by�the�range�of�vocabulary�used�in�a�social�encounter�or�in�discussing�a�topic�(the�number�of�unique�words�and�the�total�number�of�words�used).��The�more�frequently�used�words�carry�much�of�the�important�information�that�distinguishes�between�general�social�contexts.��Of�all�the�available�words�and�constructions,�some�specific�subset�is�chosen�for�use�because�it�is�needed�to�encode�that�communication.��8��� T  �� @ �Middle�range�words�carry�most�of�the�interesting�contextual�information�because�they�are�generally�known�and�used,�appear�in�different�social�contexts,�and�their�relative�use�varies�widely�from�one�social�context�to�another.��These�words�include�the�general�classes�of�nouns,�verbs,�adjectives�and�adverbs�which�allow�description�and�evaluation�across�settings.��They�also�include�(with�augmentation�from�the�top�50�or�so�words)�the�pronouns,�adverbs�and�adjectives�which�specify�and�structure�the�situation.��MCCA�focuses�particular�attention�on�the�middle�range�and�more�widely�used�words.�� @ �Contextual�information�is�also�contained�in�the�focus�upon�some�words�or�word�groups�compared�to�others.��This�can�be�seen�in�probability�distribution�patterns�across�idea/word�categories.��Individuals�use�ideas/words�in�distinctive,�patterned�ways�which�reflect�role�and�location�within�a�social�system�as�well�as�individual�socialization�and�other�individual�factors.��Sub-cultures�have�typical�overall�patterns�of�the�relative�use�of�conceptual�categories.��Furthermore,�specific�social�settings�or�contexts�appear�to�have�typical�idea/word�patterns.��Individuals�learn�these�patterns,�and�their�speech�reflects�changes�in�patterns�when�they�become�involved�in�different�social�settings�(e.g.�from�church�to�job�to�recreation�to�school).��Usage�patterns,�such�as�these,�come�to�typify�and�distinguish�institutionalized�social�settings�(Namenwirth�1968;�Cleveland�et�al.�1974).��Finally,�contextual�information�is�contained�in�the�connectedness�or�co-occurrence�of�ideas.�� @ �However�specific�or�general,�the�social�meaning�of�the�situation�is�important�because�it�provides�the�starting�point�for�individual�social�interaction.��Knowing�the�social�context�means�that�a�person�is�aware�of�what�general�kind�of�activity�is�likely�and�what,�generally,�is�appropriate�behavior.��� @ �Contextual�information�is�also�useful�in�distinguishing�between�multiple�meanings�of�certain�words,�such�as�"service",�as�illustrated�below.����'��U�X'X����*s�+, ddd Xdd Xdd X��#��#s��,�dd ��,�dd ��+  ��  C%�" ������@}}&��"Service"�� (�%V# ��(�Meanings� %�&� $ ��%�Context� %�&� % ��%�a.��Religious�gathering̀������e.g.,�church�"service"� %)�"' ��%�Traditional� %Q("( ��%�b.��Provision�of�aid,�assistancè������e.g.,�road�"service"� %(+�$* ��%�Practical� %g*$+ ��%�c.��Introductory�game�ploỳ������e.g.,�tennis�"service"� %>-�&- ��%�� �Emotional�&},,&.   &��#�'X��X�'��U�:#��� �� @ �An�empirical�approach�to�the�problem�of�measuring�social�context�proposed�here�makes�use�of�the�content�analysis�framework.��This�provides�a�basis�for�a�more�precise�evaluation�of�the�meaning�of�social�contexts�as�well�as�comparison�of�communication�across�contexts.��MCCA�attempts�to�systematically�code�contextual�information�from�textual�data.���� �A�Framework�of�Four�General�Contexts� �� � _ ��It�is�possible�to�distinguish�a�large�number�of�different�contextual�dimensions.��However,�we�have�found�it�convenient�and�useful�to�work�with�a�small,�comprehensive�set�of�four�general�contexts:�(a)�traditional,�(b)�practical,�(c)�emotional,�and�(d)�analytic.�� @ �These�four�satisfy�several�criteria�we�consider�important�for�any�set�of�contextual�"markers"�used�in�social�science�investigation.��First,�it�must�be�possible�to�integrate�the�set�of�contexts�into�social�science�inquiry�in�a�way�which�contributes�to�social�science�theory�and�interpretation.��Secondly,�they�should�broadly�address�all�aspects�of�society.��That�is,�each�context�should�substantially�contribute�to�a�kind�of�"triangulation"�which�would�help�to�locate�any�potential�text�in�relation�to�each�of�the�"marker"�contexts.��Thirdly,�the�set�of�contexts�should�contribute�to�the�interpretation�of�ideas/words�which�may�be�identified�as�important.��In�particular,�the�dimensions�should�help�identify�the�different�meanings�of�a�given�word�which�has�multiple�meanings�and�is�thus�contextually�ambiguous�(i.e.�"spring",�"service").��Fourthly,�contexts�should�contribute�to�the�explanation�of�social�behavior�by�making�it�possible�to�move�across�levels�of�analysis,�considering�individual�social�motivation�as�well�as�institutionalized�social�perspectives.��Finally,�the�set�of�contexts�should�fit�with�the�general�social�uses�of�language.�� @ �In�our�application�of�MCCA�four�scores�indicate�the�relative�closeness�of�an�analyzed�text�to�each�of�the�four�general�contexts.��These�four�scores�(called�C-scores)�are�used�as�measures�on�the�four�contextual�dimensions.��Some�evidence�suggests�that�they�can�be�used�to�define�an�approximately�orthogonal�space�(Anderson�1970;�Cleveland�et�al.�1974;�Osgood�1957),�although�this�is�not�a�necessary�criterion.��Distances�between�texts�in�this�four-space�can�be�computed�and�used�to�express�the�proximity�of�texts�to�each�other.�� @ �Each�of�the�four�contexts�incorporates�a�general�idea�of�societal�activity�and�represents�a�different�framework�within�which�specific�concepts�can�emerge:��0 @ �� � �a)�Traditional�Context.��A�normative�perspective�on�the�social�situation�predominates�and�the�situation�is�defined�in�terms�of�standards,�rules�and�codes�which�guide�social�behavior.� @�#@�# ��0 @ �� � �b)�Practical�Context.��A�pragmatic�perspective�of�the�social�situation�predominates�and�behavior�is�directed�toward�the�rational�achievement�of�goals.� @�#@�# ��0 @ �� � �c)�Emotional�Context.��An�affective�perspective�predominates�and�the�situation�is�defined�in�terms�of�expressions�of�emotion�(both�positive�and�negative),�and�maximizing�individual�involvement,�personal�concern�and�comfort.� @�#@�# ��0 @ �� � �d)�Analytic�Context.��An�intellectual�perspective�predominates�and�the�situation�is�defined�in�objective�terms.� @�#@�# �Table�1�schematically�illustrates�the�way�in�which�word�groupings�in�a�conceptual�dictionary�can�reflect�idea�categories�and�certain�idea�categories�may�be�emphasized�more�heavily�in�certain�social�contexts.��In�MCCA,�ambiguous�words�are�contextually�disambiguated�to�determine�the�most�appropriate�category,�and�each�category�receives�some�differential�weight�reflective�of�its�usage�in�the�different�contexts.� u-$'+ ��F�E*75%!p p � � `��F�� ������#��#�� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ��#��#����������&X��XX'X���� �Conducting�Contextual�Content�Analysis� �� �4 ��Conceptual�content�analysis�focuses�on�ideas�in�text.��Contextual�content�analysis�lends�perspective�to�those�ideas.��The�two�sets�of�scores�are�used�together.��There�are�a�series�of�stages�in�the�execution�of�a�research�design�incorporating�contextual/conceptual�content�analysis.��#�'X��XX&X���O#����� @ �The�first�stage�involves�the�methodological�choice�of�content�analysis.��9��A�contextual/conceptual� � �content�analysis�is�appropriate�in�one�of�three�situations.��Because�it�is�basically�an�approach�to�measurement,�it�is�useful�in�(a)�descriptive�or�explanatory�studies�--�especially�where�one�wants�to�identify�and�contrast�meanings�for�one�or�more�text�units,�(b)�in�hypothesis�testing,�or�(c)�in�exploratory�inquiries�--�especially�where�questions�are�complex,�uncharted�and�changing,�(e.g.�new�constructs,�developing�issues,�or�changes�over�time).�� @ �The�second�stage�involves�decisions�on�specific�research�procedures.��Contextual/conceptual�content�analysis�involves�all�the�usual�considerations�in�research�such�as�design,�measurement,�sampling,�pretesting,�data-gathering,�all�of�the�possibilities�of�statistical�analysis�and�reporting.��10��In�each� [$ ! �of�these,�standard�considerations�about�theoretical�grounding�and�craftsmanship�apply.�� @ �Since�contextual�content�analysis�examines�patterns�of�use�of�ideas�in�text,�it�is�important�that�the�text�qualify�as�research�data.��That�is,�it�must�be�relevant�to�the�research�problem�and�contain�characteristic�patterns�of�word�usage�rather�than�an�edited�or�altered�pattern�of�usage.��In�the�interview�situation,�skill�in�providing�a�free,�natural�stimulus�to�expression�with�minimal�intrusive�constraint�is�important.��Use�of�a�verbatim�transcript�(or�a�representative�sample�from�it)�is�critical�because�it�contains�the�pattern�information�central�to�contextual/conceptual�analysis.��11��A�machine-readable�computer�file� �*i$( �of�the�desired�verbatim�text�is�created.��Word�processors�are�useful�for�this�purpose,�and�optical�scanners�are�available�which�read�printed�text�and�convert�it�directly�into�a�computer�file.�� @ �The�third�stage�involves�the�scoring�procedures�themselves.��MCCA�uses�a�conceptual�dictionary� u-$'+ �augmented�with�the�four�contexts�(traditional,�practical,�emotional�and�analytic).��The�computer�matches�each�word�in�the�text�against�the�word�meanings�in�the�dictionary,�keeping�a�running�tally�of�usage,�concept�by�concept.��Words�not�in�the�dictionary�are�tallied�in�a�"leftover"�list.��Conceptual�category�tallies�are�percentaged�for�each�text�by�the�total�words�in�the�text.��This�score�is�subtracted�from�an�expected�score�obtained�from�a�norm�to�yield�an�emphasis�score�for�each�of�the�concepts�included�in�the�dictionary.��11��It�is�important�to�take�account�of�variability�in�the�use�of�ideas/words�across�social� � � �contexts.��This�is�done�by�dividing�by�the�standard�deviation�of�category�usage�across�the�four�contexts�to�yield�useful�emphasis�scores�(E-scores):��B�B=>1!` 0  p��B�� ������#��#�� ���������� ���������� ���������� ���������� ��#��#�����where���E��-score(��i,k��)�is�the�E-score�for�category�"��i��"�in�text�"��k��";���p(i,k)��is�the�observed�proportion�of�text�in� &�  �conceptual�category�"��i��"�for�text�"��k��";���P(i)��is�the�overall�expected�probability�of�use�of�category�"��i��";�and� �  ���S(i)��is�the�expected�standard�deviation�of�category�usage�across�contexts.��12��� ��  �� @ ���E�scores��are�computed�for�each�of�116�idea/word�categories�distinguished�in�the�current� �� �dictionary.��They�are�the�basic�measures�used�for�the���conceptual��analysis.��The�pattern�of�connectedness� �y �of�various�ideas�in�a�text�is�examined�using�a�clustering�routine.��Similarity�and�distinction�between�texts�in�terms�of�emphasized�patterns�of�ideas�can�be�quantified�as�well.��A�distance�between�texts�can�be�measured�as�a�discrepancy�between�texts�on�their�profile�of�relative�use�of�the�117�conceptual�categories�(the�117th�category�is�the�"left-over"�list�of�uncategorized�words).��The�structure�of�conceptual�differences�shown�in�this�proximity�matrix�can�also�be�examined�by�clustering�and�other�statistical�techniques.�� @ �Four���C�scores��or�contextual�scores�are�also�created�during�computer�processing�of�the�text.��As� )� �each�word�is�identified�and�classified�into�a�conceptual�category,�four�cumulative�contextual�scores�are�each�updated�as�illustrated�in�Table�3.��The�updating�uses�weights�which�reflect�the�relative�use�of�each�conceptual�category�in�the�four�general�social�contexts.��At�any�point�during�processing,�these�accumulating�scores�are�available�to�be�used�in�contextually�disambiguating�ambiguous�words.��Context�scores�are�used�to�decide�between�alternative�categorizations.�� @ �Accumulated�contextual�scores�over�a�text�are�standardized.��These�four�scores�are�the�four�contextual�dimension�measures.��Distances�between�texts�in�this�four-space�can�be�computed�and�used�to�express�the�proximity�of�texts�to�each�other,�in�terms�of�their�approach�to�the�ideas�that�are�discussed.��Cluster�analysis�helps�display�the�structure�of�this�proximity�matrix.�� @ �Finally,�E-scores�often�identify�fruitful�starting�points�for�further�qualitative�analysis.��The�computer�can�further�assist�qualitative�analysis�by�sorting�and�organizing�the�text,�by�searching�for�all�instances�of�the�use�of�some�word�or�phrase,�or�by�showing�the�use�of�key�words�in�sentences�and�phrases�in�the�text�through�key�word�in�context�(KWIC)�lists.��An�inspection�of�these�phrases�often�permits�a�refinement�of�the�sense�of�the�general�conceptual�categories�and�helps�identify�broader�concepts�extending�across�several�conceptual�categories.��This�grounding�draws�on�strengths�of�qualitative�approaches�to�text�analysis�within�a�systematic,�comparative�research�framework.���&X��XX'X���� @ �Variables,�including�composite�indices�developed�from�contextual�and�conceptual�analysis� �+Q%) �scores,�can�be�included�in�a�data�set�for�statistical�analysis,�together�with�variables�developed�in�any�of�the�other�more�traditional�ways.���We�have�treated�C-�and�E-scores�as�continuous,�interval-level� t-#'+ �measures.���������#�'X��XX&X�áh#��F�E?C5%!p p � � `��F�� ������#��#�� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ��#��#������� �An�Illustration� �� )� ��To�illustrate�the�ability�of�contextual/conceptual�analysis�to�distinguish�between�texts,�we�selected�short�passages�from�five�published�articles:��0 @ �� � �(a)�Scholarly�Journal�Article�-�report�of�research�findings�on�the�impact�of�pre-retirement�programs�on�post-retirement�satisfaction�and�behavior�for�older�people;� @�#@�# ��0 @ �� � �(b)�Airline�Finance�News�-�a�financial-page,�newspaper�report�about�an�airline's�financial�problems�associated�with�costs�of�long-distance�flights;� @�#@�# ��0 @ �� � �(c)�Magazine�Fiction�-�a�newsstand�magazine�story�incorporating�"stream�of�consciousness"�writing�about�personal�feelings,�reflections�on�life�situation,�and�future�possibilities;�� @�#@�# ��0 @ �� � �(d)�Religious�Devotional�-�a�passage�from�a�daily�religious�reading�providing�guidance�for�personal�living�(this�selection�dealt�with�feelings�of�depression�and�loneliness);�� @�#@�# ��0 @ �� � �(e)�Editorial�on�Recent�Crimes�-�an�incensed�editorial�reaction�to�a�series�of�person-crimes,�coupled�with�demands�for�action�to�be�taken�by�authorities.� @�#@�# �The�first�four�texts�were�expected�to�show�relative�emphasis�on�one�of�the�four�contexts.��The�last�text�was�chosen�to�illustrate�the�use�of�contextual/conceptual�content�analysis�in�locating�an�ambiguous�text�in�relation�to�other�texts.�� @ �The�texts�were�entered�verbatim�into�a�computer�file�and�checked�for�accuracy�and�spelling�but�not�edited�in�any�way.��Using�the�MCCA�8.3�contextual/conceptual�content�analysis�program,�each�text�was�scored�both�conceptually�(E-scores)�and�contextually�(C-scores).� u-$'+ �� @ �Table�3,�below,�provides�a�profile�of�four�contextual�scores�for�each�of�these�five�passages.��In�this�example,�the�scores�are�standardized�to�sum�to�zero�for�a�text,�to�aid�in�comparing�relative�emphasis�on�contexts.��The�higher�the�positive�score,�the�more�the�text�can�be�said�to�focus�on�a�contextual�dimension.��Greater�negative�scores�reflect�less�emphasis�on�that�context.���F�EJM5%!p p � � `��F�� ������#��#�� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ��#��#������ @ �There�are�two�major�contrasts�to�be�explored�in�Table�3.��First,�the�analyst�can�examine�distinctions�between�articles.��It�is�clear�in�the�table�that�the�scholarly�article�is�the�highest�of�all�texts�on�the���analytic���contextual�dimension.��The�magazine�fiction�and�the�crime�editorial�have�high�C-scores� �b �on�the���emotional��dimension.��Both�the�scholarly�article�and�the�airline�finance�news�have�high���pragmatic��� �K �scores�while�the�religious�devotional�and,�to�some�extent,�the�crime�editorial,�emphasize�the���traditional��.�� �4 �An�analysis�of�contextual�scores�provides�information�on�similarity�between�texts�in�the�approach�taken�to�the�topic�at�hand.��Here,�differences�clearly�reflect�what�one�might�expect�texts�from�these�sources�to�be.����� @ �The�second�analytic�path�is�an�examination�of�the�pattern�of�emphasis�on�different�contexts�for�a�given�text.��Each�of�the�pieces�is�not�exclusively�in�a�single�context�but�has�a�primary�focus�and�under-emphasizes�certain�other�contexts.��That�is,�the�text-author�opted�to�emphasize�a�given�contextual�approach�at�the�expense�of�certain�other�approaches�which�could�have�been�utilized.��Often�the�particular�pattern�of�positive�and�negative�C-scores�is�important�to�examine.�� @ �For�example,�the�scholarly�article�contains�both�practical�and�analytic�elements,�reflecting�both�the�practical�concern�of�pre-retirement�programs�as�well�as�the�analytic�approach�of�research.��Interestingly,�the�contextual�pattern�for�the�crime�editorial�emphasizes�the�emotional�component�stemming�from�the�writer's�anger�at�the�lack�of�resolution,�and�traditional�elements�about�what�"should"�be�done.�This�is�accomplished�by�de�emphasizing�practical�and�analytic�approaches.�� @ �C-scores�for�each�text�can�be�included�with�other�variables�in�a�more�extensive�analysis�of�hypothesized�causes,�consequences�or�correlates�of�the�social�context�of�texts.��Context�scores�can�also�be�used�as�controls�in�an�analysis�of�other�relationships.��A�distance�measure�(shown�in�table�4)�can�be�computed�between�texts,�using�the�four�C-scores�for�each�text�in�Table�3.��This�can�be�used�in�cluster�analyses�to�display�the�structure�of�contextual�differences.�� @ �The�five�texts�differ�in�the�way�they�approach�their�subject�matter.��The�scholarly�article��and�the�airline�financial�news�are�contextually�more�similar��quite�different�in�approach�from�the�magazine�fiction.��The�crime�editorial�falls�between�the�emotionality�of�the�magazine�story�and�the�traditionalism�of�the�religious�text.��Equipped�with�theoretical�expectations,�such�plots�have�proven�helpful�in�measuring�patterned�differences�in�context�between�texts.� t-#'+ ����� �+Q%) �����F�EHP5%!p p � � `��F�� ������#��#�� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ��#��#������ @ �Table�5�shows�the�E-score�pattern�for�the�five�illustrative�texts�on�25�of�the�116�conceptual�categories�identified�in�the�conceptual�dictionary.��Conceptual�categories�where�the�absolute�value�(or�maximum�absolute�difference�in�E-scores�across�texts)�met�a�heuristic,�substantive�criterion�(i.e.�10�or�larger)�were�selected.��The�rationale�is�that�small�differences,�approaching�zero,�are�close�to�the�overall�norm�for�use�of�that�category�and�that�important�emphasis,�omission,�or�differences�between�texts�are�of�initial�analytic�interest.�� @ �Table�5�shows�that�the�five�illustrative�texts�deal�with�quite�different�concepts�as�was�anticipated.��The�general�character�of�the�concepts�with�which�they�each�deal�can�be�identified�simply�from�the�ad�hoc�labels�of�the�dictionary's�broad�conceptual�categories.��Thus,�the�crime�editorial�dealt�with�the�need�for�sanctions�to�be�applied�for�wrongs�which�had�occurred.��The�role�of�authorities�and�what�they�should�be�expected�to�do�were�central�ideas.��On�the�other�hand,�the�religious�devotional�dealt�with�wanting�to�change�one's�life�in�some�way�to�deal�with�or�avoid�feelings�of�depression�and�loneliness.��The�magazine�story�was�toying�with�ways�of�seeing�things�and�changing�viewpoints,�themes�that�begin�to�be�reflected�in�conceptual�category�labels�in�Table�5.��Some�of�these�ideas�are�shared�by�other�texts�as�well.��Handling�more�immediate�financial�obligations�involved�in�flying�can�be�seen�in�E-scores�for�the�airline�news�text.��Finally,�some�of�the�ideas�involved�in�retiree�activities�and�the�helpful�role�of�former�employer's�programs�can�be�seen�in�the�scholarly�article.�� @ �A�proximity�matrix�can�be�computed�showing�the�distance�between�the�overall�profile�of�ideas�emphasized�in�different�texts.��This�matrix�of�"conceptual"�differences�between�texts�(not�shown),�like�the�matrix�of�"contextual"�differences�between�texts�shown�in�Table�4,�above,�can�be�analyzed�with�cluster�and�other�analytic�techniques.��Differences�and�similarities�between�texts�may�be�evident�in�the�context�of�the�discussion�and/or�the�kinds�of�concepts�which�are�discussed.�� @ �Most�analysis�would�not�stop�at�this�point�but�instead�would�make�use�of�further�detail�on�vocabulary�and�phrases�found�in�each�of�the�texts�which�accounts�for�notably�high�or�low�E-scores.��MCCA�provides�lists�of�words�from�each�text�which�were�classified�in�each�conceptual�category.��An�examination�of�these�often�leads�to�an�identification�of�the�specific�conceptual�meanings�only�broadly�hinted�in�the�overall�category�labels.��More�appropriate�theoretical�concepts�may�be�identified�from�a�pattern�of�E-scores�over�more�than�one�conceptual�category,�using�more�elaborate�measurement�models�(Weber�1983).����� �*h$( ����� �*h$( ��F�ERU5%!p p � � `��F�� ������#��#�� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ���������� ��#��#������ �The�issues�of�validity�and�reliability� �� Z$ ! ��Issues�of�validity�and�reliability�for�contextual�content�analysis�are�best�addressed�in�specific�research�situations�as�is�true�for�other�measurement�techniques�(e.g.�Likert�scaling)�where�concepts�and�their�measures�can�be�examined.��A�number�of�authors�have�addressed�reliability�and�validity�questions�in�hand�content�analysis�(Krippendorff�1980;�Holsti�1969;�Weber�1983;�Andren�1981).��There�are�a�number�of�relevant�observations�we�would�make�about�computer�content�analysis.�� @ �Important�aspects�of�reliability�in�content�analysis�are�handled�by�the�use�of�a�computer.��Computer�content�analysis�procedures�process�a�given�text�file�reliably,�in�accordance�with�instructions�in�a�specific�program.��This�is�a�major�advantage�in�that�computers�permit�investigators�to�realistically�consider�the�inclusion�of�verbatim�text�in�systematic�research�on�larger�or�more�representative�samples� u-$'+ �and�on�contrasting�varieties�of�substantive�topics.�� @ �There�are�larger�questions�of�reliability,�however,�which�need�attention,�such�as�the�reliability�involved�in�production�of�text�(e.g.�will�two�conversations�with�the�same�person�on�the�same�topic�yield�the�same�patterns,�controlling�for�pretesting�and�other�change�factors?).��There�is�also�an�interesting�area�of�inquiry�involving�sampling�variability�of�patterns�given�a�certain�number�of�respondents,�documents,�words,�topics,�and�the�like,�which�may�require�some�reconceptualization�prior�to�application�of�traditional�sampling�theory.��In�some�instances,�for�example,�it�appears�that�one�should�take�account�of�the�over-time�branching�pattern�developed�in�a�conversation�on�a�topic�or�acknowledge�that�some�conversations�amount�to�different�respondents�independently�articulating�a�widely�shared�aspect�of�culture.��In�any�event,�it�appears�from�experience�that�stability�in�patterns�is�often�reached�with�relatively�few�words�(500�to�1000�on�a�topic)�or�modest�samples�of�respondents.���� @ �It�is�generally�inappropriate�to�generalize�validity�findings�to�a�whole�measurement�approach�although�certain�typical�strengths�and�weaknesses�may�be�identified.��Few�social�science�theories�provide�a�deductive�link�with�word�patterns�and�constructs.��Instead,�investigators�using�content�analysis�rely�upon�many�of�the�same�assumptions�about�shared�meanings�which�are�central�to�the�use�of�other�measurement�approaches�(e.g.�survey�questionnaires).��The�assumption�that�there�are�wide�areas�of�consensus�about�the�meaning�of�words�appears�to�be�justified�(illustrated�by�the�usefulness�of�a�standard�dictionary)�in�many�research�situations.��There�is�also�widespread�reliance�upon�face�(content)�validity�assessment�of�measures,�which�emerges�from�informed�judgments,�prior�knowledge,�tests�of�intuitive�plausibility,�etc.�� @ �Can�anything�be�said�in�general�about�validity�of�content�analytic�measures,�including�those�computed�by�MCCA?��Our�experiences�to�date�suggest�rather�direct�links�between�content�analysis�measures�and�certain�theoretical�concepts.��One�reason�for�this�may�be�that�content�analysis�allows�one�to�operate�directly�on�the�expressed�meanings�and�emphases�of�subjects�rather�than�requiring�them�to�translate�their�experiences�into�structured�statements�closer�to�the�researcher's�needs.��The�emphases�and�nuances�of�respondents�are�preserved.��Another�possible�reason�for�a�sense�of�greater�substantive�validity�is�that�the�data�are�more�likely�to�be�gathered�as�a�part�of�the�normal�process�of�human�communication�rather�than�the�"encoded"�or�"pre-structured"�conversation.�� @ �There�have�been�a�number�of�opportunities�for�predictive�validity�checks�within�research�projects.��For�example,�a�study�which�compared�open-ended�conversation�by�husbands�and�wives�about�their�relationship,�resulted�in�accurate�classifications�of�couples�both�in�behavioral�terms�(seeking�divorce,�seeking�outside�help,�or�coping)�and�as�compared�with�independent�judgments�made�by�clinicians�with�access�to�the�couples�(McDonald�and�Weidetke�1979).�� @ �Another�study�assessed�open-ended�responses�to�a�series�of�proposed�new�financial�services�to�be�offered�by�a�bank.��The�content�analysis�data�were�able�to�accurately�predict�the�success�or�failure�of�the�actually�offered�services�(Pirro�1981).��A�recent�study�of�religious�meaning�and�life�satisfaction�(Woodworth�1982)�used�both�a�content�analytic�assessment�of�interview�transcriptions�about�life�satisfaction�and�a�widely�used�scale�item.��Initial�comparisons�suggested�a�lack�of�correspondence�between�the�scale�item�and�content�analytic�scores�but�further�examination�indicated�that�some�of�the�minor�content�analytic�scores�could�predict�the�structured�scale�rating�(R�squared�=�.86)�but�the�main�ideas�that�were�emphasized�by�respondents�in�discussing�their�satisfaction�were�in�quite�different�areas.��This�suggests�predictive�validity�of�that�which�a�traditional�scale�measures,�but�points�to�the�utility�of�being�more�sensitive�to�a�subject's�own�emphases.��A�content�analytic�assessment�of�interviews�with�sixteen�prison�ex-inmates�about�their�prison�experience�led�to�the�accurate�post-diction�of�general�type� t-#'+ �of�crime,�escape�record�and�recorded�recidivism�over�the�twenty�years�prior�to�the�interview,�all�from�a�few�content�analytic�scores�(Felt�and�McTavish�1983).�� @ �MCCA�has�also�been�utilized�in�an�examination�of�the�validity�of�scale�items.��Here,�each�scale�item�is�treated�as�a�distinct�"text"�for�which�C-scores�and�E-scores�are�derived.��A�number�of�analyses�have�been�pursued�such�as�discriminant�function�analysis�where�scale�(or�sub-scale)�discrimination�in�terms�of�the�content�analytic�meaning�scores�is�attempted.��In�suggestive�analyses�of�standard�Likert�scales�(e.g.�alienation,�job�design,�job�satisfaction),�a�small�set�of�relevant�E-scores�and�C-scores�accurately�classify�scale�items�on�their�theoretically�appropriate�scale.��Inaccurately�classified�items�reflected�scale-item�problems�discussed�in�the�literature.��Further,�the�particular�MCCA�scores�in�the�classifying�equations�reflect�the�substantive�character�of�the�concept�which�the�scale�items�were�purported�to�measure�(McTavish�and�Felt�1985;�Pierce�et�al.�1984).��A�comparison�of�published�reliability�scores�for�scales�and�the�contextual�distance�between�scale�items�suggests�that�contextual�ambiguity�may�contribute�to�scale�unreliability.��Contextual�content�analysis�was�useful�in�identifying�problem�items�and�suggesting�more�specific�ways�in�which�items�could�be�improved�to�fit�with�the�rest�of�the�scale�(McTavish�1987).��MCCA�has�also�been�useful�in�an�examination�of�organizational�context�and�role�distance�in�nursing�homes�(McTavish�and�Felt�1987).�� @ �It�has�been�helpful�to�include�traditional�measures�of�key�concepts�in�content�analytic�studies�as�well�as�separately�measured�predictors.��Adding�"criterion�text"�profiles�to�an�analysis�also�aids�validity�assessment.��If�the�criterion�text/s�represent�a�relatively�pure�instance�of�a�theoretical�construct�or�position�on�a�theoretical�continuum,�then�analysis�of�the�distance�of�other�texts�from�the�criterion�can�aid�in�analysis�and�validity�assessment.��Predictive�and�post-dictive�studies�are�useful�where�alternative�information�can�identify�relevant�criteria�to�predict.�� @ �Finally,�one�can�often�identify�expected�relationships�which�can�be�examined�in�content�analysis.��Our�experience�is�that�during�the�process�of�analysis�a�whole�series�of�relatively�low�level�expectations�can�be�generated�and�tested�along�the�line�of�"if�this�is�an�accurate�interpretation,�then�one�would�also�expect�that�to�be�true..."��The�result�is�a�series�of�small�construct�validity�tests�or�"triangulations"�which,�in�sum,�lend�substantial�insight�into�validity�questions.���� �Conclusions� �� � e ��Computer-based,�contextual/conceptual�content�analysis�augments�traditional�measurement�approaches�and�provides�a�further�means�by�which�strengths�of�qualitative�and�quantitative�research�can�be�integrated.��It�would�appear�to�expand�realistic�possibilities�for�reliable�and�systematic�analysis�of�a�broader�range�of�social�science�data�such�as�historic�documents,�cross-cultural�materials,�transcripts�of�interviews,�ongoing�verbal�processes,�and�open-ended�responses.�� @ �Further�theoretical�and�quantitative�work�is�needed�on�linkages�between�conceptual�definitions�of�key�social�science�variables�and�patterns�of�word�usage�as�well�as�on�expectations�for�comparative�word�patterns�across�cultures,�societies,�institutions,�organizations�and�historic�time.��Work�to�date�suggests�that�an�ability�to�more�directly�deal�with�expressions�of�social�meanings�in�a�rigorous�analytic�framework�is�possible�and�useful�for�social�science�investigation.�  �*i$( �� �Notes� �� Q ����'��U�X'X���1.�0 @ �We�are�indebted�to�Susan�Schrader�for�her�editorial�assistance.�$�@�#@�# �2.�0 @ �The�most�recent�introductory�summary�is�by�Weber�(1985).��Wood�(1980)�summarizes�alternatives�and�options�in� �� �computer�content�analysis.��Earlier�literature�includes�Berelson�(1952,�1954),�I.�de�Sola�Pool�(1959),�Holsti�(1969),�North�et�al.�(1963),�Gerbner�et�al.�(1969),�Krippendorff�(1980),�and�Rosengren�(1981).��Stone's�General�Inquirer�computer�program�is�a�prominent,�early�computer�content�analysis�system�developed�at�Harvard�(Stone�et�al.�1966).��Grimshaw�(1973,�1974,�1980)�reviews�some�of�the�social�science�literature�on�language.��Content�analysis�appears�in�several�social�science�methods�texts�such�as�Kerlinger�(1973)�and�its�integration�into�general�methods�has�been�discussed�(Markoff�et�al.�1975).��Recent�special�issues�of�journals�have�presented�suggestive�uses�of�microcomputers�in�handling�text�field�note�data�in�research�("Computers�and�Qualitative�Data"�in���Qualitative�Sociology��(1984);�and� ,�  ���Quality�and�Quantity��(1984).��� @�#@�# �3.�0 @ �MCCA�8.3,�the�Minnesota�Contextual�Content�Analysis�computer�program�(version�8.3�by�McTavish),�is�operating� �]  �on�a�Control�Data�Cyber�174�computer�at�the�University�of�Minnesota,�Twin�Cities.��The�package�of�programs�which�accomplishes�contextual�content�analysis�operates�directly�on�a�machine-readable,�verbatim�text�file.��Three�output�files�are�created:�a)�an�across-group�comparative�summary�of�results�of�the�MCCA�analysis�including�cluster�analyses�and�co-occurrence�analysis,�b)�a�file�of�results�and�certain�diagnostic�indices�summarized�separately�for�each�text�analyzed,�and�c)�a�data�file�of�content�analytic�scores�for�each�text�(including�C�scores�and�E�scores)�which�can�be�used�as�a�part�of�a�data�base�for�further�statistical�analysis�in�programs�such�as�SPSS.��Distinctive�word�usage�is�identified�and�these�can�be�used�as�starting�points�for�further�examination�of�the�text�(e.g.�key�word�in�context�listings�and�phrase�retrievals).�� @�#@�# ��0 @ �� � �The�analysis�reported�here�was�partially�supported�by�grants�from�Rural�Sociology�and�the�University�Computer�Center,�University�of�Minnesota.��Currently,�MCCA�8.3�is�being�extended�and�converted�for�use�on�micro�computers�and�mainframe�computer�systems�other�than�the�CDC�Cyber�at�the�University�of�Minnesota�on�which�it�is�currently�operating.��Arrangements�for�research�use�of�the�MCCA�program�on�the�University�of�Minnesota�computer�can�be�made�by�contacting�Don�McTavish,�Department�of�Sociology,�University�of�Minnesota,�Minneapolis,�MN��55455.� @�#@�# �4.�0 @ �These�include�the�Harvard�social-psychological�dictionaries�which�are�based,�in�part,�on�the�work�of�Talcott�Parsons� �� �and�embody�structural-functional�concepts�(Stone�et�al.�1966�chapters�4�and�5;�1974);�the�Lasswell�Value�Dictionary�centers�around�his�theory�of�values�(Peterson�and�Brewer�1965;�Lasswell�1968).���Special�topic�dictionaries�exist�such�as�the�Pirro�African�Dictionary�(Pirro�1968);�a�verbal�style�dictionary�(Hart�1984);�the�Institutional�Rhetorics�Dictionary�(Cleveland�1972);�and�others),�as�well�as�dictionaries�for�several�languages�including�French�and�German.� @�#@�# �5.�0 @ �Here�we�are�interested�in�an�aspect�of�the�social�use�of�language�or�"pragmatics"�(Levenson�1983;�Bates�1978;� �2! �Watslawick�et�al.�1967).� @�#@�# �6.�0 @ �More�specific�contextual�meaning�may�be�shared�by�virtue�of�participation�in�an�organization,�as�an�organizational� !�# �"culture"�(Barley�1983;�Felt�1985).��Many�more�specific�contexts�exist�as�well,�in�particular�places�or�times�(e.g.�firm�A's�mail�room�or�marketing�division).��Individuals�typically�participate�in�several�different�social�contexts�in�the�course�of�a�day�(e.g.�family,�work,�ball�game,�college�class),�shifting�language�patterns�as�they�do�so.� @�#@�# �7.�0 @ �Parsons�1951,�1964;�Rose�1962;�Stone�and�Farberman�1970;�Rochberg-Halton�1982;�Deese�1965;�Berger�and�  $�' �Luckman�1967;�Carroll�1957.� @�#@�# �8.�0 @ �In�the�U.S.�(and�apparently�in�other�Western�societies�as�well),�a�few�words,�specifically�the�most�frequently�used� �%:) �50�words,�account�for�a�high�percentage�of�word�use�(about�43�percent�for�English).��The�next�4950�most�frequently�used�set�of�words�(up�to,�the�5,000th�word)�account�for�another�48�percent�of�word�usage.��Words�beyond�the�top�5,000�or�so�account�in�the�aggregate�for�about�9�percent�of�word�usage�and�are�often�technical�terms�or�proper�nouns�which�are�restricted�to�highly�specific�conversations�or�situations.� @�#@�# �9.�0 @ �One�does�content�analysis,�of�any�variety,�when�there�is�theoretical�reason�to�believe�that�meaning�is�embodied�in� P)�". �a�set�of�written�or�spoken�textual�materials.��Additional�discussion�of�the�incorporation�of�content�analyses�as�a�method�for�research�can�be�found�in�Krippendorff�(1980),�Holsti�(1969),�Markoff�et�al.(1975),�and�Weber�(1985).� @�#@�# �10.�0 @ �Content�analysis�studies�involve�many�of�the�same�sampling�issues�that�are�involved�in�other�types�of�social�science� �+B%1 �research.��For�example,�the�selection�of�texts�to�examine�is�a�reflection�of�the�theoretical�problem�and�the�types�of�needed�contrasts.��Texts�may�be�selected�to�reflect�different�actors�or�organizations�or�time�periods,�or�a�combination�such�as�selection�of�material�over�the�course�of�a�conversation�between�two�people�where�both�speaker�and�time�are� �-�'4 �distinguished.��Content�analytic�texts�can�be�selected�from�individuals,�groups,�societies,�organizations�or�various�analytic�groupings�of�text.��Text�to�be�content�analyzed�generally�should�be�an�accurate,�unedited,�verbatim�statement.��In�some�instances,�theory�may�dictate�editing�out�irrelevant�material.� @�#@�# ��0 @ �� � �The�level�at�which�sampling�is�done�depends�to�some�extent�on�the�theoretical�level�of�interest�in�research.��Generally,�the�more�complex�the�organizational�level�to�be�investigated,�the�more�extensive�the�block�of�text�that�would�be�identified�as�the�unit�in�sampling�(e.g.�word,�phrase,�sentence,�paragraph,�document,�or�set�of�documents.).��Traditional�probability�sampling�procedures�are�relevant�and�useful�in�content�analysis.� @�#@�# �11.�0 @ �The�contextual�analysis�proposed�here�is�based�on�the�differential�profile�of�use�of�language�by�individuals�and� � G �groups�as�established�by�regularities�in�language�use�in�several�settings.��Various�works�on�language�and�its�use�have�helped�to�establish�some�of�the�regularities.��Kucera�and�Francis�(1967)�provide�word�counts�and�percentages�for�a�large�sampling�of�public,�written�material�produced�in�the�United�States�in�the�1960's.��Probabilities�associated�with�word�use�can�be�used�as�a�general�norm�or�expectation�for�use�of�written�words/ideas�in�a�broad,�contemporary�American�context.��Differences�from�such�expectations�become�significant�for�social�science�text�analysis.��A�summary�of�a�variety�of�US�based�text�tends�to�confirm�the�stability�of�these�general�probability�distributions.� @�#@�# �12.�0 @ �For�example,�in�the�sentence,�"Work�like�mine�keeps�me�from�doing�my�best.",�the�idea�of�"self"�appears�three�times� ��  �(e.g.�mine,�me,�my)�out�of�nine�words�or���p(i,k)��=�.333.��If�the�expected�occurrence�of�this�category�is���P(i,k)��=�.045� �O  �and�the�standard�deviation�of�this�category's�usage�across�contexts�is���S(i)��=�.028,�then�the�E-score�can�be�calculated� a  �as�follows:�#�'X��X�'��U��#��"� @�#@�# ��>�B-` 0  `�>�� ������#��#�� ���������� ���������� ���������� ��#��#�������'��U�X'X����0 @ �This�suggests�that�the�idea�category�"self"�occurs�more�frequently�than�one�would�expect�if�the�nine�words�reflected� �6 �usual�English�conversation.� @�#@�# ��#�'X��X�'��U��#���� �References� �� �� ����'��U�X'X����0 @ �� � �Andren,�G.�(1981).�"Reliability�and�Content�Analysis",�chapter�2�in�K.�E.�Rosengren,�(ed.),���Advances�in�Content�Analysis��.� �] �Beverly�Hills,�CA:�Sage.� @�#@�# ��0 @ �� � �Anderson,�A.�B.�(1970).��"Structure�of�Semantic�Space",�chapter�17�in�E.�F.�Borgatta�&�G.�W.�Bohrnstedt,�(eds.),���Sociological�Methodology:�1970��.�San�Francisco,�CA:�Jossey-Bass.����@�#@�# ��0 @ �� � �Barley,�S.�R.�(1983).�"Semiotics�and�the�Study�of�Occupational�and�Organizational�Cultures",���Administrative�Science� �a  �Quarterly��28:�393-413.��s "!@�#@�# ��0 @ �� � �Bates,�E.�(1978).���Language�and�Context:��The�Acquisition�of�Pragmatics��.�San�Francisco,�CA:�Academic.��4!�"@�#@�# ��0 @ �� � �Becker,�H.�S.�&�Gordon,�A.�C.�(1982).��"Field�Work�with�the�Computer�and�Criteria��for�Assessing�Programs",�paper�presented�at�the�April,�1982�Annual�Meeting�of�the�Midwest�Sociological�Society.� @�#@�# ��0 @ �� � �Berelson,�B.�(1952).���Content�Analysis�in�Communications�Research��.�New�York,�NY:�Free�Press.�w#&%@�#@�# ��0 @ �� � �Berelson,�B.�(1954).��"Content�Analysis",�chapter�13�in�G.�Lindzey,�(ed.),���Handbook�of�Social�Psychology,�volume�1��.� 8$�& �Reading�MA:��Addison-Wesley.� @�#@�# ��0 @ �� � �Berger,�P.�L.�&�Luckmann,�T.�(1967).���The�Social�Construction�of�Reality��.�Garden�City,�NY:�Doubleday.��%i(@�#@�# ��0 @ �� � �Carroll,�J.�B.,�(ed.)�(1957).���Language,�Thought,�and�Reality:��Selected�Writings�of�Benjamin�Lee�Whorf��.�Boston,�MA:�The� {&* ) �Technology�Press�of�Massachusetts�Institute�of�Technology.� @�#@�# ��0 @ �� � �Cleveland,�C.�E.�(1972).�"Institutional�Rhetorics�as�Universes�of�Discourse",�PhD�Dissertation,�Northwestern�University,�Evanston,�IL.� @�#@�# ��0 @ �� � �Cleveland,�C.�E.,�McTavish,�D.�&�Pirro,�E.�B.�(1974).�"Contextual�Content�Analysis",�Proceedings�of�the�ISSC/CISS�Workshop�on�Content�Analysis�In�the�Social�Sciences,�a�conference�sponsored�by�the�Standing�Committee�on�Social�Science�Data�of�the�International�Social�Science�Council,�UNESCO,�Centrol�Nazionale�Universitario�de�Calcolo�Electtronico�(CUNCE),�Pisa,�Italy,�September�5-13,�1974.� @�#@�# ��0 @ �� � �Conrad,�P.,�&�Reinharz,�S.�(eds.)�(1984).�"Computers�and�Qualitative�Data",�special�issue�of���Qualitative�Sociology��7:1�and� �,2&1 �2.� @�#@�# ��0 @ �� � �Deese,�J.�(1965).���The�Structure�of�Associations�in�Language�and�Thought��.�Baltimore,�MD:�Johns�Hopkins�University�Press.�.�'3@�#@�# ��0 @ �� � �Felt,�D.�(1985).�"Review�of�the�Literature�on�Organizational�Culture".�Unpublished�manuscript.�University�of�Minnesota,�Minneapolis.� @�#@�# ��0 @ �� � �Felt,�D.�&�McTavish,�D.�G.�(1983).�"Contextual�Content�Analysis�of�Interviews�with�Selected�Prison�Ex-Inmates:�A�Technical�Note",�mimeo.�University�of�Minnesota,�Minneapolis.� @�#@�# ��% �Q ��0 @ �� � �Gerbner,�G.,�Holsti,�O.�R.,�Krippendorff,�K.,�Paisley,�W.J.�&�Stone,�P.�J.�(eds.)�(1969).���The�Analysis�of�Communications� U  �Content:�Developments�in�Scientific�Theories�and�Computer�Techniques��.�New�York,�NY:�John�Wiley.� �@�#@�# ��0 @ �� � �Grimshaw,�A.�D.�(1973).�"On�Language�in�Society:�Part�I",���Contemporary�Sociology��2�(6):�575-585.�� �@�#@�# ��0 @ �� � �Grimshaw,�A.�D.�(1974).�"On�Language�in�Society:�Part�II",���Contemporary�Sociology��3�(1):�3-11.�� G@�#@�# ��0 @ �� � �Grimshaw,�A.�D.�(1980).�"Social�Interactional�and�Sociolinguistic�Rules",���Social�Forces��58�(3):�789-810.�Y @�#@�# ��0 @ �� � �Hart,�R.�P.�(1984).���Verbal�Style�and�Presidency:�A�Computer�Based�Analysis��.�New�York,�NY:�Academic�Press.� � @�#@�# ��0 @ �� � �Holsti,�O.�(1969).���Content�Analysis�for�the�Social�Sciences�and�Humanities��.�Reading,�MA:�Addison-Wesley.�� � @�#@�# ��0 @ �� � �Holsti,�O.�(1968).�"Content�Analysis",�chapter�16�in�G.�Lindzey�&�E.�Aronson,�(eds.),���The�Handbook�of�Social�Psychology,� �K  �2nd�edition��.�Reading,�MA:�Addison-Wesley.�] @�#@�# ��0 @ �� � �Holsti,�O.�R.�&�North,�R.�C.�(1966).�"Perceptions�of�Hostility�and�Economic�Variables",�pp.�169�190�in�R.�Merritt�&�S.�Rokkan�(eds.),���Comparing�Nations��.�New�Haven,�CT:�Yale�University�Press.��� @�#@�# ��0 @ �� � �Kerlinger,�F.�N.�(1973).���Foundations�of�Behavioral�Research,�2nd�edition��.New�York,�NY:�Holt,�Rinehart�&�Winston.��O @�#@�# ��0 @ �� � �Krippendorff,�K.�(1980).���Content�Analysis:�An�Introduction�to�its�Methodology��.�Beverly�Hills,�CA:�Sage.�a @�#@�# ��0 @ �� � �Kucera,�H.�&�Francis,�W.�N.�(1967).���Computerized�Dictionary�of�Present�Day�American�English��.�Providence,�RI:�Brown� "�  �University�Press.� @�#@�# ��0 @ �� � �Lasswell,�H.�D.�(1968).���Propaganda�Technique�in�the�World�War��.�New�York,�NY:�Knopf.��S@�#@�# ��0 @ �� � �Lasswell,�H.�D.,�Learner�D.�&�de�Sola�Pool,�I.�(eds.)�(1952).����The�Comparative�Study�of�Symbols��.�Stanford,�CA:�Stanford� e �University�Press.� @�#@�# ��0 @ �� � �Levenson,�S.�C.�(1983).���Pragmatics��.�London:�Cambridge�University�Press.���@�#@�# ��0 @ �� � �Lindkvist,�K.�(1981).�"Approaches�to�Textual�Analysis",�chapter�1�in�K.�E.�Rosengren,�(ed.),���Advances�in�Content�Analysis��.� �W �Beverly�Hills,�CA:�Sage.� @�#@�# ��0 @ �� � �Markoff,�J.,�Shapiro�G.�&�Weitman,�S.�R.�(1975).�"Toward�the�Integration�of�Content�Analysis�and�General�Methodology",�chapter�1�in�D.�R.�Heise,�(ed.),���Sociological�Methodology:�1975��.�San�Francisco,�CA:�Jossey-Bass.���@�#@�# ��0 @ �� � �Martel,�M.�U.�(1968).�"Age-Sex�Roles�in�American�Magazine�Fiction�(1890-1955)",�chapter�5�in�B.�L.�Neugarten,�(ed.),���Middle�Age�and�Aging:�A�Reader�in�Social�Psychology��.�Chicago,�IL:�University�of�Chicago�Press.�m@�#@�# ��0 @ �� � �McDonald,�C.�&�Weidetke,�B.�(1979).�"Testing�Marriage�Climate",�MA�theses,�Iowa�State�University,�Iowa�City,�IA.� @�#@�# ��0 @ �� � �Naisbett,�J.�(1982).���Megatrends:�Ten�New�Directions�Transforming�Our�Lives��.�New�York,�NY:�Warner.���@�#@�# ��0 @ �� � �Namenwirth,�J.�Z.�(1968).�"Contextual�Content�Analysis",�mimeo.�Yale�University.� @�#@�# ��0 @ �� � �Namenwirth,�J.�Z.�&�Lasswell,�H.�D.�(1970).���The�Changing�Language�of�American�Values:�A�Computer�Study�of�Selected� q   �Party�Platforms��.�Beverly�Hills,�CA:�Sage.�2�!@�#@�# ��0 @ �� � �North,�R.�C.,�Holsti,�O.�R.,�Zaninovich,�M�G.�&�Zinnes,�D.�A.�(1963).���Content�Analysis:�A�Handbook�with�Applications�for� ��" �the�Study�of�International�Crisis��.�Evanston,�IL:�Northwestern�University.�� c#@�#@�# ��0 @ �� � �Ogilvie,�D.�M.,�Stone,�P.�J.�&�Schneidman,�E.�S.�(1966).�"Some�Characteristics�of�Genuine�Versus�Simulated�Suicide�Notes",�chapter�16�in�Stone�et�al.���The�General�Inquirer:�A�Computer�Approach�to�Content�Analysis��.�Cambridge,�MA:�MIT� 6"�% �Press.� @�#@�# ��0 @ �� � �Osgood,�C.�E.,�Suci,�G.�J.�&�Tannenbaum,�P.�H.�(1957).���The�Measurement�of�Meaning��.�Urbana,�IL:�University�of�Illinois� �#g' �Press.� @�#@�# ��0 @ �� � �Parsons,�T.�(1951).���The�Social�System��.�Glencoe,�IL:�Free�Press.�:%�)@�#@�# ��0 @ �� � �Parsons,�T.�(1964).���Social�Structure�and�Personality��.�Glencoe,�IL:�Free�Press.��%�*@�#@�# ��0 @ �� � �Peterson,�R.�&�Brewer,�T.�(1965).�"The�Laswell�Value�Dictionary",�mimeo.�Yale�University.� @�#@�# ��0 @ �� � �Pierce,�J.,�McTavish�D.G.�&�Knudsen,�K.�R.�(1986).�"The�Measurement�of�Job�Design:�A�Content�Analytic�Look�at�Scale�Validity",���Journal�of�Occupational�Behavior��7:�299�313.�>(�!-@�#@�# ��0 @ �� � �Pirro,�E.�B.�(1968).�"A�Computer�Analysis�of�African�Political�Ideology:�Ghana,�Guinea",�PhD�Dissertation,�Yale.� @�#@�# ��0 @ �� � �Pirro,�E.B.�(1981).�"Financial�Study�of�West�Coast�Banking",�unpublished�manuscript.�Iowa�State�University,�Ames.� @�#@�# ��0 @ �� � �Pirro,�E.�B.�&�McTavish,�D.G.�(1982).�"Contextual-Conceptual�Dictionary,�version�one",�mimeo,�&�"version�two",�in�preparation.�University�of�Minnesota,�Minneapolis.� @�#@�# ��0 @ �� � �Pool,�I.�de�Sola,�(ed.)�(1959.)���Trends�in�Content�Analysis��.�Urbana,�IL:�University�of�Illinois�Press.�,�%2@�#@�# ��0 @ �� � �Rochberg-Halton,�E.�(1982).�"Situation,�Structure�and�the�Context�of�Meaning",���The�Sociological�Quarterly��23:�455-476.��,s&3@�#@�# ��0 @ �� � �Rose,�A.�(ed.)�(1962).���Human�Behavior�and�Social�Processes��.�Boston,�MA:�Houghton�Mifflin.��-4'4@�#@�# ��0 @ �� � �Rosengren,�K.�E.�(ed.)�(1981).���Advances�in�Content�Analysis��.Beverly�Hills,�CA:�Sage.�Q@�#@�# ��0 @ �� � �Sankoff,�G.�(ed.)�(1980).���The�Social�Life�of�Language��.�Philadelphia,�PA:�University�of�Pennsylvania�Press.��@�#@�# ��0 @ �� � �Schlagheck,�D.�(1985).�"Contextual�and�Conceptual�Content�Analysis�in�the�Study�of�Foreign�Policy�Decision-Making",�a�PhD�dissertation,�University�of�Minnesota,�Minneapolis.� @�#@�# ��0 @ �� � �Stone,�G.�P.�&�Farberman,�H.�J.�(eds.)�(1970).���Social�Psychology�Through�Symbolic�Interaction��.�Waltham,�MA:�Xerox� U  �College�Publishing.� @�#@�# ��0 @ �� � �Stone,�P.�J.,�Dumphy,�D.�C.�Smith,�M.�S.�&�Ogilvie,�D.�M.�(1966).���The�General�Inquirer:�A�Computer�Approach�to�Content� � � �Analysis��.�Cambridge,�MA:�MIT�Press.�� G@�#@�# ��0 @ �� � �Watzlawick,�P.,�Beavin,�J.�H.�&�Jackson,�D.�D.�(1967).���Pragmatics�of�Human�Communication��.�New�York,�NY:�W.�W.� Y  �Norton.� @�#@�# ��0 @ �� � �Weber,�R.�P.�(1983).�"Measurement�Models�for�Content�Analysis",���Quality�and�Quantity��17:�127-149.�� � @�#@�# ��0 @ �� � �Weber,�R.�P.�(1985).���Basic�Content�Analysis��.�Beverly�Hills,�CA:�Sage.��K @�#@�# ��0 @ �� � �Wood,�M.�(1980).�"Alternatives�and�Options�in�Computer�Content�Analysis",��Social�Science�Research��9:�273-286.�] @�#@�# ��0 @ �� � �Woodworth,�D.�L.�(1982).�"A�Content�Analytic�Study�of�Religious�Meaning�and�Life�Satisfaction",�PhD�Dissertation,�University�of�Minnesota,�Minneapolis.� @�#@�# �