Revisiting Inference. A Multidisciplinary Approach

TABLE OF CONTENTS

0. Introduction
0.1. Background to the study
0.2. Rationale behind the study
0.3. Major Claims
0.4. Research Objectives
0.5. Research Questions
0.6. Structure of the Dissertation

CHAPTER ONE: LITERATURE REVIEW
1.1. Formal Logic and the Study of Inference
1.1.1.What is Logic? Towards a definition
1.1.1.1. Boundaries between logic and other disciplines
1.1.1.2. The relevance of Logic
1.1.2.The Logical Study of Inference
1.1.2.1. Is logic all about inference?
1.1.2.1.1.The matter and form of logic
1.1.2.1.2. The Matter and Form of Inference
1.1.2.2. The contributions of logic to the study of inference
1.1.2.2.1. Delimitating the boundaries of inference
1.1.2.2.2. Types of Inference
1.1.3.Types of Logic and Inference Analysis
1.1.3.1.Deductive Logic
1.1.3.1.1. On validity
1.1.3.1.2. Categorical Logic
1.1.3.1.2.1. Basic Notions
1.1.3.1.2.2. Validity Testing
1.1.3.1.2.3. Criticism of categorical logic
1.1.3.1.3.Truth-functional logic
1.1.3.1.3.1. Basic Notions
1.1.3.1.3.2.Validity Testing
1.1.3.1.3.3.Criticism of truth-functional logic
1.1.3.2.Inductive logic
1.1.3.2.1.Inductive syllogism
1.1.3.2.2.Inductive generalizing
1.1.3.2.3.Inductive analogy
1.2. Pragmatics and the Study of Inference
1.2.1. Inference in Major Pragmatic Theories
1.2.1.1. Inference in Grice’s Paradigm
1.2.1.1.1. Possible behaviours towards the maxims
1.2.1.1.2 Revisiting the debate over the inferentiality of implicatures
1.2.1. 2. Inference in speech act theory
1.2.1.2.1. Speech Act theory
1.2.1.2.2. Inferencing about speech acts as a type of utterance interpretation
1.2.2. Inference at the interface of Pragmatics, Semantics, and Logic
1.2 2.1. Inference at the interface of Pragmatics and Semantics
1.2.2.1.1. Introducing Presupposition
1.2.2.1.2. Semantic vs. pragmatic presupposition
1.2.2.2. The Pragmatics/Logic Interface
1.2.2.2.1. Informal Logic
1.2.2.2.2. Argumentation Schemes
1.3. Conclusion
1.3.1. Summary of literature review
1.3.2. New Inference Taxonomy

CHAPTER TWO: METHODOLOGY
2.1. Rationale behind methodological choices
2.2. Corpus
2.2.1. Corpus description
2.2.2. Corpus selection criteria
2.2.3. Corpus collection method
2.3. Analytical Method

CHAPTER THREE: CORPUS ANALYSIS
3.1. Qualitative Analysis: text-by-text inferential description
3.1.1. Qualitative Analysis of d-texts
3.1.2. Qualitative Analysis of g-texts
3.2. Quantitative Analysis
3.2.1. The Charts
3.2.2. Quantitative Findings
3.3. Discussion
3.3.1. Text-by-text interpretation of the findings
3.3.2. General Discussion

CONCLUSION

References

APPENDICES
Appendix A: List of Argumentation Schemes
Appendix B: Corpus

ACKNOWLEDGEMENTS

I would like to express my deepest gratitude to Professor Mounir Triki whose incessant encouragment and expert guidance have helped me persevere in this academic work.

I should also like to express my heartful gratitude to my husband Hafedh for his unfailing moral support and technical assistance.

I am also gratefully indebted to my father Mohsen and mother Samira for their constant presence, to my in-laws Hachemi and Majida for their support, and to all my friends and colleagues for their encouragement and readiness to help.

Abstract

Previous literature on inference has often approached it from a single perspective offering but an incomplete understanding of its full potential. For this reason,the study proposes a multidisciplinary approach to inference. Synthesizing the contributions of Logic and Pragmatics, this research builds a new Inference Taxonomy. The study claims that this newly developed theoretical model can be transformed into a discourse descriptive tool meant to complement traditional tools offered by Discourse Analysis. The most important contributions of the research is that it attempts to offer a new understanding of inference, develop that theoretical understanding into a new instrument that serves to describe discourse, and gauge the functioning of that instrument.

In order to test the functioning of the new descriptive tool and to explore the insights it can offer, the study makes use of a corpus-based approach. It applies theoretical and methodological findings to a corpus of ten texts. The analysis consists in providing a systematic description of the different types of inference underlying the texts, which is congruent with the inherently qualitative method used. Quantification is also used with the purpose of investigating the implications of the qualitative analysis.

The corpus analysis illustrates how the new inference-based instrument can provide detailed descriptions of the inferential structure of texts. It equally sheds light on the correlations that may exist between inferences found in a text and the type or topic of this text, possible similarities between different texts in inferential terms, clear tendencies or common inferential features, etc. Thus, the findings substantiate the research major claims.

The outcome of the study is a fine-tuned multidisciplinary inference taxonomy that is translatable into an instrument of descriptive discourse analysis, which can be optimized to cover comparative, evaluative, critical, and pedagogic ends.

Lists of tables, charts, and figures

Tables

Table 1: Possible behavioural cases towards the maxims

Table 2: Corpus Description

Table 3: Corpus Sources

Table 4: Corpus Abbreviations

Table 5: Inference Indicators (Cogan, 1998, p. 5)

Charts

Chart 1: Made Inferences

Chart 2: Triggered Inferences

Chart 3: Inference Ratios

Chart 4: Presumptive inferences

Figures

Fig.1: Logical validity of arguments used in Carnap’s example

Fig.2: Other valid arguments having the same logical form as example (8)

Fig. 3: Another argument having the same logical form as example

Fig. 4: standard-form translation of a syllogism into a categorical syllogism

Fig. 5: Venn diagrams of standard-form categorical claims (borrowed from Moore & Parker, 2009, p. 257)

Fig. 6: Pragma-semantic inferences

Fig.7: New Inference Taxonomy

Fig. 8: Sample Results Page

Fig. 9: Example of text

Fig. 10: Note in Text

Fig. 11: The New Inference Taxonomy Revisited

0. Introduction

This introductory section sets itself the following tasks: explicating the background to the study, evincing the overarching rationale underpinning the whole work, pronouncing the main claims, describing the major objectives, putting forward the questions to be addressed, and providing an overview of the whole dissertation.

0.1. Background to the study

The idea of conducting the present research emanated from the desire to answer some of the interrogations raised by a previous piece of research which dealt with the concept of "enthymemes"(Ben Amor &Rapatel, 2008). One of the main findings of that research is that the enthymeme, an inference type, has been for long the subject of theoretical controversies and misconceptions, and that it needs to be revisited from a multidisciplinary perspective.

In the same vein, and as an attempt to enlarge the scope of that research, the current study intends to examine the concept of inference, the hypernym of enthymeme (Reed & Walton 2005; Walton, 2001; Vega, 2000; Pelletier, 1996; Sorensen, 1988), and to see whether a multidisciplinary approach to inference could shed more light on such a concept. The notion of inference has been dealt with by a wide range of disciplines such as psychology and logic (Johnson-Laird, 1983; Salmon, 1967; Reade, 1938). However, it has not received enough attention by linguistics, except semantics and pragmatics where it was studied under different labels, namely under the notion of entailment in semantics (Cruse, 2000) and implicatures in pragmatics (Cummings, 2005). For that reason, the present research claims that linguists, and more specifically discourse analysts, should benefit from other disciplines' contributions to gain a deeper insight into the functioning of such a concept. Therefore, this research aims to pursue a study that takes into consideration previous contributions to the understanding of inference, with the objective of investigating the potential of inference as a descriptive discourse analytic tool, then utilizing the findings in the analysis of a corpus encompassing a variety of text genres.

0.2. Rationale behind the study

Inference is "a human act" (Reade, 1938, p. 4) that deserves deep scrutiny and critical enquiry since it is a fundamental concept in the life span of our beliefs. Indeed, "[i]nferences are important for three reasons: (1) They can cause or generate beliefs. (2) They can support beliefs. (3) They can justify beliefs." (Lambert & Ulrich, 1980, original emphasis, p. 30). Another reason that motivates the choice of inference by this research is that it has been the subject of a theoretical controversy. On the one hand, it is argued that inference governs all human behavior ranging from the simplest perceptions to the most elaborate demonstrations (Leighton in Leighton & Sternberg, 2004; Manktelow& Over, 1990; Johnson-Laird, 1983). On the other hand, an opposite view restrains the meaning of inference to conscious and deliberate acts of reasoning (Nagao, 1990; Salmon, 1967; Reade, 1938). Between the two extremes which claim that "all knowledge is inferential and that there is no such thing as inference" (Reade, 1938, p. 4), the present study suggests a compromise that takes into consideration aspects of both views.

Moreover, inference deserves investigation because of its close link with knowledge. “[I]nference is essential for our communication with each other and the development of everyday, as well as advanced scientific, knowledge” (Manktelow& Over, 1990, p. vii). Whenever we need knowledge, we need to infer; or, as Reade puts it, "in so far as we know already, we do not need to infer" (ibid.). Inference is present in every instance of knowledge acquisition and development. In this respect, Johnson-Laird (1983) states that "to define inferential thinking is to define life" (p. 23). Therefore, inference is an important but unclear concept which deserves to be revisited from new angles, i.e.; multidisciplinary ones.

As stated above, the study of inference has witnessed two opposite currents which claim that "all knowledge is inferential and that there is no such thing as inference" (Reade, 1938, p. 4).This difference in defining the scope of the same concept results from the difference in perspectives. Most of those defending the first opinion are cognitive linguists, whereas the other view is held mainly by logicians. Johnson-Laird, however, acknowledges that inference, as a function performed by the human mind, is “too complicated to be seen clearly, or to be studied with advantage, from the perspective of a single discipline” (Johnson-Laird, 1983, p. xi). Therefore, a multidisciplinary approach is needed in order to define inference while taking on board all relevant contributions made by the concerned disciplines. Accordingly, the present research proposes a distribution of these disciplines into two study axes: a logical axis, and a pragmatic axis.

Both logic and pragmatics, as hereby utilized, are not referred to as curbed and independent disciplines. Rather, they are used to refer to whole areas of study in which the disciplines per se have a central position along with a wide scope of neighboring concepts, emerging theories, and borderline notions. In other words, logic and pragmatics are used in their umbrella dimension and not as well-defined, self-restrained arenas.

Apart from the undeniable fact that both logic and pragmatics have made considerable contributions to the study of inference (see 1.1.2. and 2. for more details), their selection as fundamental theoretical foundations of the present research emanates from a central assumption underlying this work, namely the assumption that inference is reasoning and discourse . Though visible and amenable to study at the level of discourse, inference is symptomatic and revealing of a deeper level; that of reasoning.

0.3 Major Claims

The present dissertation is governed by two major claims: one theoretical and the other is practical. On a theoretical level, a comprehensive understanding of inference requires the recourse to different disciplines. The study, therefore, claims that inference has to be revisited from a multidisciplinary approach in order to be grasped in its full potential.

On a practical level, the research claims that inference, and more precisely, the new understanding of inference envisaged by this study, has important implications on our ability to describe and analyze discourse. In other words, the research claims that inference can be seen as a new discourse analytic tool; since analysing discourse in inferential terms can contribute a wealth of information, such as the degree of explicitness of the speaker’s reasoning, the predominant inferential types and subtypes used by the speaker, the speech acts carried by a given chunk of discourse (both directly and obliquely), the assumptions that the speaker has about her/his audience, the existent correlations between the inferential choices depicted in a text, and the text genre, etc.

The present study, therefore, posits that already existing tools of descriptive discourse analysis have to be complemented by a new instrument which concerns itself with inference description, because to describe a text on the inferential level can bring about insights that cannot be otherwise elucidated. To sum it up, the overarching claim motivating this study is that inference, if studied from a multidisciplinary perspective, can be developed into a novel and insightful discourse descriptive tool.

0.4 Research Objectives

By dividing the dissertation into three chapters, the current study purports to realize three main objectives. The objective of the literature review part is to review the contributions of different disciplines to the study of inference in view of synthesizing a multidisciplinary understanding. The objective of the methodology part is to develop the synthesized understanding into a new descriptive toolkit that would allow the discourse analyst to describe the inferential net underlying any piece of discourse. Last but not least, the main objective of the corpus analysis part is to test the validity of the new toolkit by applying it to a corpus of diversified texts.

This research situates itself within the realm of descriptive discourse analysis (as opposed to critical discourse analysis) and, therefore, contents itself with the task of constructing the new descriptive tool and testing its validity. Consequently, evaluating the strength or soundness of inferences is beyond its scope. This, however, does not deny the fact that the findings of the intended description should be the starting point for any evaluative or critical endeavour. The descriptive tool that the research aims to construct can be eventually used by researchers belonging to different disciplines (history, journalism, forensic linguistics, critical discourse analysis, etc.), for non-descriptive purposes. However, as far as the current study is concerned, the targeted end is to synthesize the descriptive tool and test its functioning by recourse to a multidisciplinary approach.

0.5 Research Questions

The present study raises and tries to answer a number of questions which are tackled in accordance with the general structure of the dissertation. The first chapter in which the literature is reviewed attempts to answer the following questions:

- Whatisinference?
- Why does it have to be understood from a multidisciplinary perspective?
- What does the multidisciplinary approach add to the previous understanding of inference?
- How do Logic and Pragmatics contribute to the understanding of inference?
- How can these contributions be synthesized into a new multidisciplinary approach to inference?

The second chapter, methodology, seeks to answer the following questions:

- How can the multidisciplinary understanding of inference be translated methodologically into a multidisciplinary toolkit applicable to the analysis of a corpus?
- What motivates the choice of the corpus?

Finally, the last chapter, the corpus analysis tries to answer questions such as:

- How is inference manifested in every discourse type?
- How is inference manifested in every theme?
- What are the outcomes of the analysis? Are there correlations, differences and similarities, etc.?
- How can these outcomes be interpreted?
- How do the results of the study support the major claims?

0.6 Structure of the Dissertation

The dissertation comprises three chapters, namely the Literature Review, the Methodology, and the Corpus Analysis. Chapter one, Literature Review, aims at reviewing the major contributions of Logic and Pragmatics to the multidisciplinary understanding intended by the present study, which justifies its composition of two sections. The first section in this chapter, entitled ‘Formal Logic and the Study of Inference’, opens on a tentative definition of Logic, a delimitation of its boundaries, and a demonstration of its pertinence to the current research. Afterwards, the section moves on to review the contributions of Logic in general to the study of inference before specifying the inputof the two types of logic: deductive logic and inductive logic. More specifically, two schools of deductive logic are to be reviewed, namely categorical logic and truth-functional logic with a detailed exposition of the tools offered by each to test the validity of inferences. Then the researcher deals with Inductive logic and exposes the different types of inductive inferencing.

The second section constituting the Literature Review chapter concerns itself with Pragmatics in its multidisciplinary dimension. It starts by explaining standard pragmatic theories in their relation to the topic of the research. More precisely, the Cooperative Principle and Speech Act Theory are to be introduced in an attempt to benefit from their approach to the concept of inference. Afterwards, the section focuses on the interface between pragmatics and other disciplines, such as semantics, in the investigation of typically inferential phenomena such as presuppositions. The research also examines the interface of pragmatics and Informal Logic, which offers practical tools in the analysis of presumptive inferences, namely argumentation schemes. The literature Review chapter has ended with a conclusion summarizing the whole review and presenting its major findings.

The methodology chapter seeks to design a detailed analytic method that would translate the theoretical findings of the Literature Review into a step-by-step process that can be applied on a corpus.For that purpose, the chapter starts by giving the rationale behind the different methodological choices made by the researcher. Another section is, then, devoted to the description of the corpus and the clarification of why and how it is selected. The method per se is, then, explained in detail, with special emphasis on the tools that are not reviewed in the Literature Review chapter.

The last chapter analyses the corpus by following the method displayed in the methodology chapter and applying it systematically to each one of the ten texts. The researcher starts by giving a brief presentation of each text (title, number of words, etc.) then proceeds to perform the steps which constitute the levels of analysis. At a later stage, the findings of the text-by-text qualitative description are gathered in charts in order to allow for the quantitative analysis to be carried out. The chapter closes on an interpretation of the findings and a description of the results. This last section is expected to clarify whether the results of the investigation support the claimed pronounced in this introduction.

CHAPTER 1: LITERATURE REVIEW

The literature review part seeks to present the notion of inference from the perspectives of logic and pragmatics. Apart from reviewing relevant literature, the motivating objective of this part is to grasp the different but related facets of inference and, consequently, develop a new method in approaching it.

1.1. Formal Logic and the Study of Inference

This chapter seeks primarily to review the contributions of logic to the study of inference. In order to do so, however, it will start by defining logic in the first place. The chapter has three sections: the first one attempts to define the discipline and show its relevance to the current investigation. Thesecond section, entitled The Logical Study of Inference, locates the concept of inference within the realm of logic, and determines the areas where logic has made essential contributions to the understanding of inference. The third and last section of this chapter reviews different types of logic and the tools proffered by each one of them for the study of inference.

1.1.1. What is Logic? Towards a definition.

This section tries to delimit the scope of logic by defining its subject-matter and spelling out its boundaries with neighboring disciplines and areas of study. The section ends by shedding light on the relevance and significance of logic to the present research.

Logic is defined as "the science of the forms of valid argument" (Turner, 2006: 87). Similar definitions stipulate that logic is a field that "studies the structure of arguments, and is primarily concerned with testing arguments for correctness or validity" (Malmkjaer, 1991, p. 129). In fact, it is this intimate connection between logic and the study of arguments that makes the recourse to this discipline necessary in the present study of inference. In other words, the fact that logic studies arguments and that arguments are linguistic "records of inferences" (Lambert & Ulrich, 1980: 30) makes it clear that logic is unavoidably relevant if we are to pursue a comprehensive study of inference. Besides, the nature of the assumptions made in the introduction, specifically the idea of the existence of a tacit reasoning level behind the apparent discourse level, necessitates the use of logic since "logic is conceived as the branch of philosophy that studies the truth and falsity of judgment, reasoning (i.e. the sequence of judgments), and the scientific method" (Auroux, 1990, p. 1495, my translation).

Considering the above-mentioned definitions, it can be advanced that logic is concerned with the study of arguments, which are representations of inferences, from a structural point of view. Logic studies the structure and not the content of inferences; it does not study the way inferential arguments function in argumentative contexts,nor the way they can be confirmed or weakened, or the reasons why they succeed or fail in persuading people, etc. Logic examines the way inferences are structured out of propositions with a view to studying their logical validity. In other words, logic does not study the content of arguments¹ but is solely "concerned with the design or structure of arguments" (Turner, 2006: 88). Logic is not concerned with the truth or falsity of the ‘factual sentences’ from which arguments are made, but “with relations between factual sentences (or thoughts)" (Carnap in Lambert & Ulrich, 1980: xv).

To adopt the definition of logic as the formal study of arguments and their validity is to accept Carnap's assertion that "logic is competent to determine whether our assertions or suppositions are consistent with one another"(Carnap in Lambert & Ulrich, 1980: xviii). If a conclusion is drawn from assumptions (premises) with which it is not consistent, this conclusion would be logically invalid. Therefore, "[t]he task of logic may also be viewed as making evident the consequences of a given assumption, irrespective of its truth or falsity"(Carnap in Lambert & Ulrich, 1980: xviii). Carnap gives the following example in order to show that logic is useful in judging arguments' consistency:

I assume that iron does not float on water and that my latch-key is made of iron, the supposition that my latch-key will float on water is incompatible with my original assumptions. In order to avoid contradicting my own premises, I must therefore assume that my key will sink. Logic itself does not affirm this last assumption; it simply renders explicit what is implicitly contained in the two premises previously assumed. (Carnap in Lambert & Ulrich, 1980: xviii)

It is not clear how logic can determine that the conclusion contradicts one of the premises without considering their content. In order to see the relevance of Carnap's example, the assertion that logic does not give importance to the content of arguments has to be understood in the sense that logic is not concerned with the truth or falsity of premises, but rather with the validity of the argument i.e. whether the conclusion follows from the premises. (A more precise definition of the concept of validity will be introduced later under section 1.1.3.1.1.). The following example builds on Carnap's to show how arguments can be valid while containing false premises.

ARGUMENT 2

Abbildung in dieser Leseprobe nicht enthalten

Fig.1: Logical validity of arguments used in Carnap’s example

Even though the first premise and the conclusion of Argument 2 in Figure 1 are false, the argument is surely valid because its conclusion logically follows from its premises. To recapitulate, the subject-matter of logic is the study of arguments' structures which allows the evaluation of their validity and consistency.

1.1.1.1. Boundaries between logic and other disciplines

The boundaries between logic and other disciplines may have become obscured by two main facts: the shift of some logicians towards other disciplines, and the difference between the original and the new areas of logical investigation. Some philosophers who had adhered to the idea that logical analysis can, by itself, account for the meanings of everyday language changed their views later. For example, Ludwig Wittgenstein, in his early writings, "adhered to logical atomism, and argued that philosophy can resolve the problems of our everyday understanding by breaking meaning down into logical atoms" (Turner, 2006: 92). Later, Wittgenstein shifted to the side of those who believe that "[m]eanings cannot be detached from how language is used in particular situations" (ibid.), and that the structures of social groups are more important than the logical properties of sentences in the understanding of human language. Therefore, it is clear that Wittgenstein, in spite of being considered as a logician because of his early claims (Huisman, 1993), is more of a pragmatician, a sociolinguist, or an anthropologist than a logician.

In addition to the shift of certain logicians towards other disciplines while still being considered as logicians, the boundaries between logic and other disciplines may have become blurred by the difference between the original and new areas of logical investigation. Since all the definitions of logic revolve around the notion of arguments and since logic's "original area of concern [is] argumentative discourse" (Britannica vol. 23 p. 225), the boundary between logic and rhetoric is not clear-cut. Turner (2006), in his review of the historical development of logic, maintains that "[l]ogic from Aristotle to the 17th century was concerned with the public nature of argument, and hence it had close connections with rhetoric, politics and the public sphere" (p. 90). Both rhetoric and logic investigate arguments but not in the same way. The latter studies the structure of arguments while the former deals with their content with a view to investigating their persuasive potential. "Logic from the 19th century was separated from the arts of public conduct and conjoined with science and mathematics" (ibid.). However, it is to be noted that the difference between 'old logic' and 'new logic' is not one of content but rather one of form. Actually, the main novelty that logic witnessed is its use of more abstract forms of representation, namely mathematical symbols, which are needed in order to "develop artificial languages such as symbolic logic, truthtables, set theory or propositional calculus" (Turner, 2006: 90). Although it was exposed to this transformation, logic is still primarily concerned with the validity of arguments but, instead of studying this validity in real arguments, which belong to the public sphere, it focuses on abstract languages which use mathematical symbols. "The aim was to purify logic of its dependence on social content, that is, of ordinary language"(ibid.). After this shift, the borderline between logic and rhetoric has become clearer. But can the logical analysis of a given argument be of any use if not complemented by a rhetorical analysis?

The logical analysis is certainly useful in determining the validity of arguments and in understanding how arguments are built out of inferences and inferential chains. But when we are to study the persuasive potential of arguments (i.e. inferential instances) in a given chunk of any type of discourse (scientific discourse, legal discourse, media discourse, etc.), knowing whether an argument is valid or invalid is not of great usefulness. In fact, in many cases, invalid arguments—like the case of logical fallacies—prove to be efficiently persuasive (Walton, Reed, and Macagno, 2010; Tindale, 2007; Hamblin, 1970). Rhetoric is therefore needed to complement the logical study of inference, thus multi-disciplinarity becomes a necessity.

Being concerned with reasoning as a type of human behaviour, logic does not have clear frontiers with other disciplines studying human behaviour. Psychology, for example, studies human reasoning as well, but from a totally different perspective. Whereas logic studies the validity of inferences, psychology is rather interested in answering questions about the mental processing of inference: how does the human mind understand inference? How does it generate inference? Is all human thought inferential? It is, according to Carnap (1980), the task of the logician to point out invalid reasoning but it is the task of the psychologist, among others, to trace the origins of such irrationality. “The logician by himself has no remedy to offer, and must turn to psychologists and social scientists for aid” (Carnap in Lambert & Ulrich, 1980: xix). In other words, logic defines logical thinking in absolute terms independently of the human subjects who conduct that thinking and of the social and discursive contexts where the thinking occurs. “The laws of human conduct in observing and violating the requirements of logical thinking must be discovered by psychology and the social sciences” (ibid.). Therefore, the task of logic, in the present study, is to define inference in “pure” terms. It will answer questions about the nature of inference, its types, the difference between valid and invalid inferences, etc. Then, logic, in its account for inference, needs to be complemented by other linguistic disciplines which are more conscious of the social dimension of language, one of which is pragmatics.

1.1.1.2. The relevance of Logic

As argued above, logic is “concerned with human behavior”(Carnap in Lambert & Ulrich, 1980: xv), but not in the same way that other sciences, such as brain physiology and psychology, are. Logic studies human reasoning as an abstract process regardless of the subject who performs it; of the circumstances and variables that affect and are affected by such a process; and of the truth or falsity of its constituent elements. Accordingly, human reasoning is seen as sets of premises and conclusions ordered into sequences, the validity of which (and not the truth or falsity) is the main focus of logic. In a nutshell,"[t]he major task of logic has been to establish a systematic way of deducing the logical consequences of a set of sentences” (Britannica vol. 23 p. 225).

It seems reasonable to maintain that logic, by itself, provides only an incomplete picture of human reasoning, a part of which is inference. This claim does not seek to diminish the role of logic or to overlook it. On the contrary, it suggests that logic is a pertinent and even necessary tool in the study of inference (as an instance of human reasoning), but whose importance can be grasped more easily when looked at from a multidisciplinary perspective. Indeed, logic is, to a considerable extent, complementary to other disciplines (such as pragmatics) in the study of arguments, i.e. records of inferences.

In real everyday discursive contexts, a great deal of human linguistic interaction is predominantly inferential in nature (Johnson, 2000; Hamblin, 1970;Toulmin, 1958), and“some inferences are justified, but others are not, and sometimes people make mistakes, or commit fallacies, in the inferences they perform” (Manktelow& Over, 1990, p. 63). Linguistics, including rhetoric, pragmatics, semantics, and discourse analysis, cannot account for the instances where “the question of which forms of inference are justified naturally arises” (ibid.). This is clearly the territory of Logic, the discipline whichaddresses such anissue. But, then, other questions which Logic cannot answer, or at least cannot answer by itself, arise, such as the following: how are logically un/justified inferences linguistically formulated? Which inferences have a performative role? How do they carry that role? Is there any correlation between the predominant logical description of inferences in a given chunk of discourse and the discursive genre or type to which the chunk belongs?

Logic cannot definitely give a comprehensive account of the nature and functioning of inference in its discursive usage. In order to preserve its purity, this discipline transcends earthly areas of study and remains in the province of abstraction. That is why it is hardly of any use unless it is complemented by more practical, down-to-earth disciplines.Therefore, the objectives set by this study necessitate the recourse to logic as well as linguistics in its broad sense.

It is to be noted also that the nature of the corpus to be analyzed in the third part of this research not only justifies the recourse to logic but shows that we have no other option than using logic in the analysis of a major part of the corpus. In fact, some of the subjects dealt with in the corpus (e.g. gene mapping) are too technical for the researcher, which would affect the verification of their truth and the evaluationof the quality of their conclusions.

In the absence of being able to verify a conclusion directly, we can evaluate how our premises (or beliefs) entail the conclusion of interest. In other words, we can learn to scrutinize our claims to truth. Our claims to truth are the processes by which we generate conclusions. […]. In the end, we may not be able to control our ability to directly verify a conclusion’s truth, but we may be able to control the cohesion or soundness of the reasoning that lead up to it (Leighton, in Leighton & Sternberg eds. 2004, p.5)

Still, philosophical and intangible as it may seem, logic is necessary in the present study of inference in that it allows us to “distinguish between thinking which is irrational or illogical and thinking which is reasonable or logical, and thus win a richer understanding of the ways in which logical and illogical thought may influence the activities of men" (Carnap in Lambert & Ulrich, 1980: xv). Thus, the relevance of logicbecomes undeniably obvious.

1.1.2. The Logical Study of Inference

This section is divided into two subsections: The first one tries to locate inference within the territory of logic while the second goes deeper into the input of logic in the process ofinference understanding.

1.1.2.1. Is logic all about inference?

Reade (1938) discusses the place of inference within the realm of logic and thereby criticizes the claim that “logic ignores the matter of reasoning (or of propositions) and studies only its form" (Reade, 1938, p. 37-38). He deals with two main questions. First, the question of whether logic is all about inference is tackled through a comparison between inference and logic in terms of matter and form. While attempting to address this issue, Reade deals with the second question of whether logic is interested solely in the form of arguments. Accordingly, the following will be divided into two subsections that address the dichotomy of matter and form in logic and inference respectively.

1.1.2.1.1. The matter and form of logic

Aristotle, who invented “the famous antithesis of form and content”(Reade, 1938, p. 36), could not apply it to logic, or at least, “he does not explicitly examine the respective parts of form and matter in logic as such” (ibid.), hence the complexity of such an issue. Nevertheless, Reade seems to be sure that logic does have a matter, and refuses the idea that logic is not concerned with truth. He asserts that though logic has no concern, unless quite incidentally, with the truths belonging to other sciences, it has a vital interest in the truth proper to itself. Assuredly, too, it must have a subject-matter of its own, for a science that has no 'matter' in the Aristotelian sense cannot possibly be said to exist.(Reade, 1938, p. 37)

The question that needs to be addressedat this stage is ‘what is this matter?’ Obviously, and as advanced by Reade, logic has to have a matter of some kind if it is to be considered as a science. In attempting to answer this very question, one may fall into confusion for two main reasons. First, “the ‘matter’ may turn out to be nothing else but ‘form’” (Reade, 1938, p. 37). As a matter of fact, and as explained earlier in this chapter and in the introduction, logic studies the form and structure of reasoning. For this reason, one may be lost in endless sequences of circular theoretical questions like: What is the matter of logic? If it is form, then what is its form? Second, the arena of logic is sometimes “so defined as to exclude all subjects but inference” (ibid.). This claim is a source of confusion because if logic is all about inference, then they would have the same matter, which is inconceivable since inference “is merely one of the subjects investigated by logic"(ibid.). To dive into such philosophical issues is far beyond the scope and the purpose of this research. Therefore, by way of conclusion to this debate, it may be ascertained that logic is a science in its own right that purports to study the form of reasoning. However, it should be added here that the fact that logic studies form does not imply that “logic has nothing to do with truth” (Reade, 1938, p. 37). This point will be explained more thoroughly under the next heading.

1.1.2.1.2. The Matter and Form of Inference

Similarly, inference is surrounded by confusion and misunderstanding as far as its matter and form are concerned. One of the above-mentioned sources of confusion is still valid here. Indeed, the idea that inference is the sole subject-matter of logic is at the origin of the claim that “inference has no matter distinguishable from its form" (Reade, 1938, p. 39). The interest of logic in the form of reasoning entails that the propositions involved in an inference are better reduced to letters or symbols since the logician is indifferent to the content of terms. Stated differently, the logical analysis of an inference, which lies mainly in determining its validity, can be done without reference to the content of the premises. It follows that if inference can be reduced to its abstract form then the matter of inference lies in its form. Reade conducts a lengthy demonstration where he chains up many arguments against the latter claim. He started his reasoning by an explanation of the case of ‘conversion’² which cannot be undertaken without reference to the content of terms. Indeed, some logic textbooks claim that “the A proposition, in the form ‘All S is P ’, can only be converted per accidens into ‘Some P is S ’” (Reade, 1938, p. 39). Reade strives to prove that one cannot say whether a case of conversion is right or wrong without knowing the nature of the relation that links terms. The nature of this relation, in its turn, cannot be determined unless the nature of the terms themselves is known.

In order to corroborate his analysis, Reade adds the example of syllogisms³ where the content of terms cannot be overlooked. In this case, "the reason […] for recognizing syllogism […] as a distinct form of inference is that it operates within the province mapped out by a special kind of relation, commonly described as the relation of subject and attribute” (Reade, 1938, p. 41). Accordingly, one has to check whether the relation that links the terms of the syllogism’s first premise is a subject-attribute relation of the sort “All humans are mortal” or “all women are female” or “all living beings that give birth are mammals”. Therefore, the abstraction of an inference into letters and symbols sometimes prevents the logician from pursuing a genuine analysis of a syllogistic inference. Actually, one has to determine the character of the main relation in an inference, and, “[t]he character of a relation is not properly intelligible without reference to the character of the terms that it relates”(Reade, 1938, p. 42). Thus, the content of the propositions included in an inference can by no means be overlooked if one is to determine whether that inference is a syllogism or not. Consequently, an understanding of the matter and form of inference begins to take shape in the way explained in the following: […] the 'matter' of inference, consisting primarily in relations, is thus implicated in questions about the nature of terms. […]. I may be allowed, perhaps, to take it as an established opinion that syllogism (pointing back, no doubt, to a certain view of substance) is a form of inference appropriate only to the relation of subject and attribute, so that propositions such as ' A is equal to B ', ' A follows B ', ' A is underneath B ', can have no place in syllogistic reasoning. (Reade, 1938, p. 43)

In order to prove that the character of a relation cannot be properly determined without reference to the character of the terms that it relates, Reade adds to his argument the case of equality. Indeed, equality is a relation according to which two sets of items prove to be equal. But, an “indispensable condition of mathematical equalization is the equivalence of units” (ibid.). The abstraction of terms into symbols or letters permits a uniformity of units and, thus, mathematical equalization becomes possible. However, “in so doing we are in fact imparting to the terms the character which alone can make it possible to express the relation of equality” (ibid.). Therefore, a deep scrutiny of the case of syllogisms shows the importance of the content of terms in reaching a proper logical understanding of inference.

In more concrete terms, Reade proposes to compare two arguments dealing with the issue of kinship: “(1) ‘If A is the ancestor of B, and B the ancestor of C, then A is the ancestor of C’: (2) ‘If A is the father of B, and B the father of C, then A is the father of C.’” (Reade, 1938, p. 44). The two arguments have exactly the same form that assures transitivity between terms, but, the first one is valid while the second is not. This judgment about their validity emanates from the logician’s knowledge of “the nature and relations of kinship, which constitute the ‘matter’ of all inference belonging to the genealogical tree” (ibid.). Therefore, the matter of inference is a relation.

Hence, two conclusions can be drawn from Reade’s arguments. First, inference does have a matter and a form which are distinguishable but inseparable. In fact, the matter of inference which consists in a relation (such as subject-attribute relation, predication, equality, kinship, etc.) is distinct from its form (examples of which are the syllogism, the aforteriori argument, conversion, etc.). However, in valid inferences, form depends on matter (the nature of the relation that links the terms of the inference). In other words, a valid argument can be said to be a syllogism, for example, if and only if, the relation governing the terms of the inference is a subject-attribute relation. The third conclusion to be drawn from Reade’s demonstration is that logic, even if it is not primarily interested in the content of inferences, cannot totally ignore content. In actual fact, and as stated earlier, the main concern of logic is the validity of inferences which largely depends on form. But form, in turn, depends on matter, which is nothing but the relation governing the inference’s terms, and the nature of a relation cannot be properly determined without reference to the nature of terms themselves. For example, how can a relation be considered to be one of kinship without knowing the (human) nature of the terms involved?

To conclude the whole section, it can be said that inference is one of the major subjects studied by logic but not the only one since the matter of logic is different from the matter of inference, the former being about the validity of human reasoning while the latter is about the relations that link the terms involved in an inference.

1.1.2.2. The contributions of logic to the study of inference

This section attempts to define inference through the delimitation of its boundaries with other neighboring concepts before presenting different inference types and typologies. It is accordingly divided into two subsections.

1.1.2.2.1. Delimitating the boundaries of inference

Within the realm of logic, inference is in the vicinity of a number of neighboring concepts, which may cause confusion. In order to define inference, a cautious delimitation of its boundaries has to be established. Indeed, some reasonably precise definitions of the concepts in question have to be introduced in order to see the difference, as well as the link, between inference and such concepts.

Inferences, being made up of premises and conclusions, can mistakenly be considered as arguments. This confusion can be explained by the fact that inferences and arguments, though different, are closely related. Lambert and Ulrich (1980) elucidate this point by saying that "inferring is a mental activity" (p. 30)whereas “[a]rguments are unlike inferences because theyare not activities of any kind; rather they are abstract things, composed of propositions” (ibid.). They further explain that arguments, “when expressedverbally or in writing, they can be records of inferences” (ibid.). It follows that, since it is a mental activity, an inference can well exist without the existence of any argument. But whenever there is an argument, it means that there has necessarily been an inference of some kind. The following clarifies more the connection that exists between inferences and arguments:

An argument represents an inference if its premises consist of the propositions that someone begins by considering (or believing), and its conclusion is the proposition that he ends by inferring consciously or unconsciously from those premises. An argument records an inference if it represents that inference and someone expresses it in writing with the intention thereby of representing the relevant inference. (Lambert & Ulrich, 1980: 30)

In other words, an argument is said to represent an inference when the inference is still an abstract thought, and to record it when the inference is spelled out, “expressed verbally or in writing” (ibid.).

Another concept with which inference can sometimes be confused is the concept of implication. Similarly, implication is different from, but closely related to, inference. As stated earlier, inference is an activity performed mentally by humans (and, to a second degree, by computers in contexts of artificial intelligence); whereas, “[i]mplication is not an activity of any sort; it is a relation between propositions. In particular, it is that relation obtaining between the premises and the conclusion of a valid argument." (Lambert & Ulrich, 1980: 30-31). Reade also tackles this issue by advancing that“by the former term [inference] is signified the human act by which we claim to advance in knowledge, by the latter [implication], the existence of some ordered system which is unaffected by the business of ‘learning’ "(Reade, 1938, p. 10). By “unaffected by the business of learning”, Reade means that whether a human being learns or does not learn, infers or does not infer, an implication can still hold. Stated differently,“[i]mplication holds (or fails to hold) independently of any human activity” (Lambert& Ulrich, 1980, p.31). Therefore, it can be advanced that human beings infer conclusions from propositions whereas propositions imply conclusions; and the conclusion inferred by the human mind may or may not correspond to the conclusion implied by the propositions. Actually, a person infers a certain conclusion from given propositions because s/he believes (correctly or mistakenly) that the given propositions imply the inferred conclusion.

Even though fundamentally different, “the activity of inference and the relation of implication” (ibid.) are intimately connected. Indeed, Lambert & Ulrich explain that “[w]here S is a person, B a propositionand A a proposition or set of propositions, it would be correct for S to infer B from A if and only if A implies B” (ibid.). This is the reason why the concept of logical implication is correlated with the notion of validity which roughly means correct reasoning (a thorough discussion of validity will be introduced in section 1.3.1.). “An inference is valid if and only if its premises logically imply its conclusion” (Manktelow& Over, 1990, p. 66).This intimate connection would lead to the conclusion of the present discussion about the boundaries between inference and argument, on the one hand, and inference and implication, on the other hand. The following statement encapsulates the connection between the three concepts: “An inference is correct if and only if the premises of the argument recording it imply the conclusion” (ibid. p. 33). Stated differently, the correctness of an inference depends on the validity of the argument recording it, which depends on whether the propositions do imply the conclusion or not.

To sum up, it can be claimed that inference, argument, and implication are three facets of the same thing. If looked at as an instance of human mental activity, human reasoning that can turn out to be correct or wrong, then it is an inference; if looked at as the concrete written or verbal manifestation of that mental activity into a full-fledged sequence of premises and a conclusion (that can be valid if the inference is correct, or invalid if the inference is wrong), then it is an argument; andif looked at as an abstract relation holding between a set of propositions and a conclusion, then it is an implication.

One final distinction that has to be clarified is the one between ‘inference’and ‘reasoning’. “Reasoning is broadly defined as the process of drawing conclusions. Moreover, these conclusions inform problem-solving and decision-making endeavors” (Leighton, in Leighton & Sternberg eds. 2004, p.3). This definition of reasoning may lead one to think that reasoning and inference denote the same thing. Advocates of this idea claim that every type of human reasoning starting from mere perception to elaborate scientific theorizing is inherently inferential and that it is“no more possible to define inferential thinking than to define life » (Johnson-Laird, 1983, 23). Peirce, for example, “estimated that every sequence of thoughts is inferential in nature” (Grize, 1996, p. 105, my translation). Critics of this stance think that, important and central to human thought as it is, inference is only one species of activities involving drawing conclusions i.e. one species of reasoning. “Almost everything we do and think involves drawing conclusions. When we learn, criticize, analyze, judge, infer, evaluate, optimize, apply, discover, imagine, devise, and create, we draw conclusions from information and from our beliefs” (Leighton, in Leighton & Sternberg eds. 2004, p.11, my emphasis). A different viewpoint is held by Walton and explained in the following quote:

It is important to distinguish between reasoning and argument. Reasoning is a sequence of inferences. An inference is a set of propositions and a Toulmin warrant joining the premise propositions to the conclusion. Reasoning is not necessarily goal-directed it can go in any unanticipated direction. But reasoning is used in argument. An argument is the use of reasoning in a dialogue exchange between two parties – a proponent and a respondent – where the joint purpose of the exercise is to realize a goal for a certain type of conversation by collaboratively reasoning together. Thus argument is essentially a dialectical notion involving two parties who reason together (Walton, 1997, p. 605)

The quote above focuses on both the sequential and purposeless nature of reasoning. The latter is actually defined as “a sequence of inferences” that “can go in any unanticipated direction”. Therefore, Walton seems to be close to the claim made by Leighton above that inference is rather a part of reasoning and not all of it. He further explains that the main difference between ‘reasoning’ and ‘argument’ lies in the fact that an argument is reasoning in a dialectical context.

1.1.2.2.2. Types of Inference

Having clarified the difference between inference and other close concepts, the present research attempts to understand inference further. In order to do so, the current subsection reviews the ways in which inference has been classified into types.

Deduction versus Induction: the whole story

The way in which inference has been classified has often taken the form of dichotomies. Oneway of looking at inference is to see it from the perspective of the dichotomy of demonstrative versus non-demonstrative inference. The former is exclusively made of deductions. The latter, however, is not exclusively made of induction, though it is often taken to be so and though induction is considered to be the antonym of deduction. "The distinction between valid deduction and non-demonstrative inference is completely exhaustive. Take any inference whatsoever, it must be deductive or non-demonstrative" (Salmon, 1967, p. 20).

It would be helpful at this point to start by defining both deduction and induction and explaining the difference between them. Deduction is defined as “any process of reasoning by means of which one draws conclusions from principles or information already known” (p. 180). In other words, “if a piece of reasoning is intended to establish the conclusion, given the truth of the premises, then whether or not it succeeds in doing so, it is said to be deductive reasoning” (Lambert & Ulrich, 1980: 4).Therefore, “the central idea here is that in a valid deductive argument the truth of the premises guarantees the truth of the conclusion; in some sense the conclusion is already contained in the premises” (p. 181). On the other hand, "if a piece of reasoning is notdeductive, but nevertheless is intended to provide some evidence for the conclusion, then it is said to be inductive reasoning."(Lambert & Ulrich, 1980: 4). Indeed, the main difference between these two types of reasoning is that, while deduction is intimately linked to validity, induction displays degrees of support to the conclusion instead of absolute validity or absence of it. Salmon (1967) explains this phenomenon as follows:

In deductive logic, correctness is an all-or-nothing affair. Deductive inferences are either totally valid or totally invalid; there cannot be such a thing as degree of validity. In inductive logic the situation is quite different. Inductive correctness does admit of degrees; an inductive conclusion may be more strongly supported than another. (Salmon, 1967, p.16)

Apart from validity, another concept that differentiates between induction and deduction is that of necessity. While validity is used to logically describe inferences, necessity concerns conclusions. Indeed, “conclusions derived deductively are considered necessary because they contain the same amount of information as that found in the premises leading up to it” (Leighton in Leighton & Sternberg ed. 2004 p. 5). There follows from the necessity associated with deduction a claim that deduction is fitter⁴ for the scientific enterprise than induction since it is “one way to yield true conclusions without checking them against reality: If the premises are true, then the conclusion drawn must be true” (ibid.). In contrast, an inductively drawn conclusion is not necessary in the sense that “it is likely to be true but it is still not necessarily true” because “unnecessary conclusions represent information that goes beyond the information already contained in the premises” (ibid. p. 5-6). The facts that inductive conclusions have more content than the premises and that deductive conclusions share the same content as their premises lead to Peirce’s dichotomy of ampliative versus explicative inference (Peirce, 1955). The latter exclusively includes deductive inferences while the former includes mainly but not exclusively inductive inferences. In fact, ampliative inference, defined as an inference that “has a conclusion with content not present either explicitly or implicitly in the premises" (Salmon, 1967, p. 8), includes, apart from inductive inference, “inference from one or more descriptions of an individual case to some further descriptions of that case”⁵ (Oxford Companion to The Mind, p. 361). Some examples would be useful in illustrating the difference between induction and deduction:

Example (5): deductive argument

All bodies existing in the atmosphere of the earth are attracted to the ground

My body exists in the atmosphere of the earth

Therefore, my body is attracted to the ground

Even if the conclusion of Example (5) is not mentioned explicitly in the argument, it can be understood from the premises. Indeed, as argued above, the informational content carried by the conclusion of a deductive argument does not exceed the content expressed in the premises. For that reason, it is rather common to come across deductive inferences that are not fully expressed, i.e., whose conclusions are left implicit. The conclusion of example (5) is necessary in the sense that, because both premises are true, one does not need to check the truth of the conclusion against reality. This phenomenon of necessity is otherwise referred to as ‘truth-preservation’ or ‘warrant-transfer’ ((Boghossian& Williamson, 2003, p.226). Once the premises are proved true and the inference form valid, the truth is preserved in the move from premises to conclusion by virtue of the valid pattern of inference able of taking the reasoning entity “from truth to truths” (Boghossian& Williamson, 2003, p.226).

Example (6): inductive argument

Plants x and y are in my garden

Plants x and y need sunny weather in order to grow green

Therefore, all plants in my garden need sunny weather in order to grow green

The informational content of the conclusion in example (6) exceeds that of the premises. In other words, the content of the conclusion is not included implicitly in the premises. Moreover, the conclusion is not necessary in the sense that the truth of the premises does not guarantee that of the conclusion. Indeed, even in cases in which both premises are true, the truth of the conclusion still has to be checked against reality since it brings more information than that which is mentioned in the premises.

A less technical way of looking at the induction/deduction dichotomy is through the well-known ascending/descending ‘directional metaphor’ explained in the following:

Within logic and philosophy, deduction is contrasted with induction. Frequently, the contrast is made by use of a directional metaphor: by induction one moves from particular to general and from the less general to the more general, ascending the theoretical ladder which terminates in first principles; by deduction one moves from more general to less general and from general to particular, descending the theoretical ladder which terminates in facts about particular individuals or events. (The Oxford Companion to The Mind, p. 180)

Though widely spread among students and even in introductory logic textbooks, the above correlation between induction and specific-to-general reasoning, on the one hand, and between deduction and general-to-specific reasoning, on the other, is quite simplistic and may be misleading at times. In fact, “The differencebetween deduction and induction has nothing to do with general or specific reasoning, but has everything to do with what the conclusion does on the basis of the premises” (Allen, 2004, p. 90). Let’s look at an example where the reasoning, though inductive, goes from the general to the specific.

Example 7:

Most politicians are liars

Mr. X is a politician

Mr. X is a liar

Clearly, this inference proceeds from the general to the specific. But is it deductive or inductive? Does the truth (hypothetical truth, of course) of the premises guarantee that of the conclusion? The answer is no. Nevertheless, it gives support for the conclusion and the strength of this support depends on the percentage of ‘politicians’ who are ‘liars’. The inference carried by the argument in example 7 is, therefore, inductive⁶.

This leads us to examine the meaning of induction more thoroughly. While deduction is characterized by the necessity of its conclusions, “conclusions derived inductively are considered unnecessary because they contain more information than that found in the premises leading up to it” (Leighton, in Leighton & Sternberg eds. 2004, p.5). As a matter of fact, inductive reasoning yields conclusions that are both unnecessary and ampliative (containing more information that the premises). Another aspect that characterizes induction in comparison to deduction is its gradable strength. While a deductive inference is either valid or invalid, an inductive inference does allow conclusions to enjoy different degrees of strength since “in an inductive argument, unlike deduction, if the premises are true, then theconclusion is only probably true and how big a chance that it is true depends on the weight of evidence presented in the premises” (Allen, 2004, p. 91). In a nutshell, deductive inference is demonstrative, valid, necessary, and explicative while inductive inference is non-demonstrative, unnecessary,ampliative, and gradable in terms of strength.

The difference between induction and deduction seems to be obvious and easily traceable. This may be true on a theoretical level but, practically speaking, when dealing with real discourse, the analyst is faced by a whole “muddy mass of indistinct cases which are, by and large, the everyday reasoning we use.” (Allen, 2004, p. 90). It has been found by some logicians that rarely do induction and deduction operate separately. Most of the time,“towards the end of a complex argument, the reasoning will become deductive, carefully delineating a logical set of relationships that, in the earlier parts of the complex argument have been established through inductive reasoning. » (Allen, 2004, p. 92). It is one of the objectives of the third part of this study (the corpus analysis part) to check the veracity of such claims and to scrutinize the use and functioning of these two types of inference in real discourse.

The discussion on inference types involves two opposing views on the superiority of either induction or deduction. The advocates of the view that deductive inference is fitter for the scientific enterprise capitalize on the necessity and truth preservation inherent in this type of inference. According to this view, necessity is important in scientific reasoning for two main reasons. First, it establishes undeniable conclusions since “[i]t is difficult to deny a necessary conclusion unless its premises are also denied”(Leighton in Leighton & Sternberg ed., 2004 p. 6). In contrast, when it comes to inductive arguments“you might be more critical of the conclusion once you discover that is only a possible conclusion, and you might therefore not change your behavior drastically in light of it” (Leighton, in Leighton & Sternberg eds. 2004, p.6). Second, this very deductive necessity guarantees the argument’s cohesion since “knowing a conclusion’s necessity […] indicates the cohesion of the reasoning that underlies it” (ibid.). In other words, if an argument is deductive, then it is valid and its conclusion necessary; and if the conclusion is necessary, then it is consistent with the premises, i.e., it does not contradict them, hence the cohesion of the whole reasoning. This is important especially in cases where it is difficult to verify the conclusion against reality. In fact, when the argument is deductive, checking the truth of the premises against reality would suffice to establish the truth of the conclusion since deduction is inherently truth-preserving. These are the reasons why advocates of this view think that “deduction is the sole mode of inference in science” (Salmon, 1967, p. 21) whereas induction has no role to play in it. They even claim that inductive logic is “a complete delusion”(ibid.) and that “the crude induction exhibits a complete absence of scientific understanding" (Salmon, 1967, p. 17) and associate it with pre-scientific reasoning. On the other hand, partisans of the second view, i.e. the superiority of induction as far as scientific reasoning is concerned, think that the “basic property of necessary truth preservation” intrinsic to deductive inference is a curse rather than a blessing:

But now we must face the grim fact that valid deductive inference, although necessarily truth-preserving, isnonampliative. It is impossible to deducefrom accepted basic statements any conclusion whose content exceeds that of the basic statements themselves. Observation statements and deductive inference yield nothing that was not stated by the observation statements themselves. If science consists solely of observation statements and deductive inferences, then talk about theories, their falsifiability, and their tests are empty. The content of science is coextensive with the content of the statements used to describe what we directly observe. There are no general theories, there is no predictive content, there are no inferences to the remote past. Science is barren. (Salmon, 1967, p. 23-24)

Salmon argues that induction is not only fitter for science than deduction, but, more importantly, science cannot even exist without induction. Nevertheless, deduction sometimes seems to bring new knowledge because, in some cases, the move from premises to conclusion“is unanticipated and therefore psychologically surprising, but the conclusion cannot augment the content of the premises” (ibid., p. 8). If science were not ampliative, it would be useful only to describe observable facts, and “would be useless for prediction, post-diction, and explanation” (ibid., p. 20).

To sum up, deduction is stronger than induction in the sense that deductive inference is valid and its conclusion is necessary and truth-preserving. Conversely, induction, which is inherently ampliative, is more fruitful than deduction in which “the content of the premises is equal to the combined factual content of all valid deductive conclusions of these premises” (ibid. p. 33). Therefore, deduction is logically more secure but is far from being sufficient for the generation of new knowledge.

Other typologies

The quote mentioned above (at the beginning of the previous section) and restated below gives the impression that Salmon views inference as a dichotomy. "The distinction between valid deduction and non-demonstrative inference is completely exhaustive. Take any inference whatsoever. It must be deductive or non-demontsrative" (Salmon, 1967, p. 20).However, Salmon, later, develops non-demonstrative inferences into two subtypes as explained in the following:

The category of non-demonstrative inferences, as I have characterized it, contains, among other things perhaps, all kinds of fallacious inferences. If, however, there is any kind of inference whose premises, although not necessitating the conclusion, do lend it weight, support it, or make it probable, then such inferences possess a certain kind of logical rectitude. It is not deductive validity, but it is important anyway. Inferences possessing it are correct inductive inferences. (Salmon, 1967, p. 8)

What Salmon argues for is that inferences can be classified into three types:demonstrative truth-preserving inferences, non-demonstrative fallacious inferences, and non-demonstrative but correct inductive inferences. There exists, as he puts it, “an exhaustive trichotomy of inferences: valid deductive inference, correct inductive inference, and assorted fallacies” (ibid. p. 10). Another way of classifying inferences is that of Nagao, (1990) who divides inferences into two types: mathematical inference “where there is no ambiguity” (p. 12) and everyday inference “where we use various compromises” (ibid.). Then, he argues that everyday inferences encompass five subtypes: probabilistic inference, inductive inference, common sense reasoning, qualitative reasoning, and analogical reasoning (p. 12-13). This division of inference into mathematical and everyday reasoning stresses the idea that inference is sensitive to genre. A similar classification of inference types is that introduced by Allen (2004) whereby he presents five reasoning categories: causal reasoning, reasoning from generalization, reasoning from specific cases, reasoning from analogy, and reasoning from terms.

[...]

---

¹ The term ‘argument’ is used here, and in the rest of the dissertation, to refer to ‘inferential argument’, i.e. an argument that records an inference, an argument that contains at least one premise and a conclusion. This note is deemed important because there are other types of non-inferential arguments. Therefore, ‘argument’ and ‘inferential argument’ are used interchangeably in this context (and not in absolute terms). Later, in section 1.2.2.1, the exact difference between an argument and an inference will be elucidated.

² “You find the converse of a standard-form claim by switching the positions of the subject and predicate terms” (Moore & Parker, 2009, p.265).

³ A syllogism is a deductive argument made of two premises. A famous example of syllogisms is “All men are mortal; Socrates is a man; Therefore, Socrates is mortal”. A more thorough explanation of this type of inference will be introduced under section 1.3.1.2.1.

⁴ The superiority of deduction over induction shall be dealt with in more details below.

⁵ This type of inference is labelled by Peirce as ‘abduction’. He considers it as a vital form of inference (Peirce, 1955)

⁶ It is more technically called ‘inductive syllogism’. More on this concept under section 1.3.2.1.